{ "data": { "similar": [ { "grade": "B", "json": "https://www.anchorterminal.com/tools/axolotl.json", "name": "Axolotl", "score": 64.8, "shared": [ "finetune.sft", "finetune.lora", "finetune.export" ], "slug": "axolotl" }, { "grade": "C", "json": "https://www.anchorterminal.com/tools/fireworks-fine-tuning.json", "name": "Fireworks AI Fine-tuning", "score": 59, "shared": [ "finetune.sft", "finetune.lora", "finetune.export" ], "slug": "fireworks-fine-tuning" }, { "grade": "C", "json": "https://www.anchorterminal.com/tools/together-fine-tuning.json", "name": "Together AI Fine-tuning", "score": 54.7, "shared": [ "finetune.sft", "finetune.lora", "finetune.export" ], "slug": "together-fine-tuning" }, { "grade": "D", "json": "https://www.anchorterminal.com/tools/unsloth.json", "name": "Unsloth", "score": 51.5, "shared": [ "finetune.sft", "finetune.lora", "finetune.export" ], "slug": "unsloth" }, { "grade": "D", "json": "https://www.anchorterminal.com/tools/tinker.json", "name": "Tinker", "score": 51, "shared": [ "finetune.sft", "finetune.lora", "finetune.export" ], "slug": "tinker" }, { "grade": "B", "json": "https://www.anchorterminal.com/tools/vertex-ai-tuning.json", "name": "Vertex AI Gemini tuning", "score": 64.2, "shared": [ "finetune.sft", "finetune.lora" ], "slug": "vertex-ai-tuning" } ], "tool": { "slug": "nebius-token-factory-fine-tuning", "name": "Nebius Token Factory fine-tuning", "vendor": "Nebius", "vendorUrl": "https://nebius.com", "kind": "http-api", "category": "fine-tuning", "summary": "Nebius Token Factory runs supervised fine-tuning jobs on open models such as Llama, Qwen, gpt-oss, Gemma and DeepSeek through an OpenAI-compatible REST API, with LoRA or full weights and downloadable checkpoints.", "url": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning", "markdownUrl": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/nebius-token-factory-fine-tuning.json", "repo": "https://github.com/nebius/token-factory-cookbook", "license": "Proprietary service (cookbook examples MIT)", "transports": [ "http" ], "remoteUrl": "https://api.tokenfactory.nebius.com/v1", "packages": [], "auth": "api-key", "authNotes": "`Authorization: Bearer` with an API key created by a person in the console under API keys, shown once and read from `NEBIUS_API_KEY` in the docs. Keys belong to a project. Project Admins and Members both have full access to the Files and Fine-tuning APIs, and no per-key scopes or expiry were found in the reviewed documentation.", "pricing": "usage", "pricingNotes": "Pay as you go, with no monthly fee stated. No fine-tuning price was found in the docs or in the public catalogue at `/api/public/models_info`. The product page sends readers to the Token Factory console's prices page, a script-drawn console page that robots.txt disallows, so we did not read it. A bank card is mandatory at onboarding. New accounts get $1 of trial credit valid for 30 days (https://docs.tokenfactory.nebius.com/other-capabilities/billing-new.md). Dedicated endpoints bill while one or more replicas are ready.", "priceSummary": "Pay per use", "where": "hosted", "x402": { "level": "no", "evidence": "No x402, MPP or L402 in the docs index, the OpenAPI file or the billing page (checked 2026-10-08).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-10-08" }, "docsUrl": "https://docs.tokenfactory.nebius.com/post-training/overview", "llmsTxt": "https://docs.tokenfactory.nebius.com/llms.txt", "openapi": "https://api.tokenfactory.nebius.com/openapi.json", "capabilities": [ "finetune.sft", "finetune.lora", "finetune.export" ], "tags": [ "hosted", "usage-priced", "card-required", "open-weights", "llms-txt", "openapi", "async-jobs" ], "lastRelease": "2026-09-30", "graded": true, "anchor": { "graded": true, "score": 47.7, "grade": "D", "agentReady": false, "rank": 738, "ranked": true, "rankOf": 842, "categoryRank": 9, "methodology": "0.4", "run": "2026-10-01", "scores": { "ergonomics": 51, "maintenance": 61, "payments": 0, "reliability": 45, "schema": 68, "security": 52, "transparency": 79 }, "pending": [ "performance", "tasks" ], "breakdown": [ { "key": "reliability", "name": "Reliability", "weight": 16, "effectiveWeight": 20, "score": 45, "points": 9, "reason": "Scored with the hosted lines. status.nebius.com is a Statuspage for the whole Nebius cloud with a Token Factory component in each of nine regions and none for fine-tuning (15 of 20). From 10 July to 8 October 2026 two major incidents were tagged Token Factory, a network incident of 93 minutes on 27 July and overheating hardware in us-central1 from 19 to 20 August lasting about 21.5 hours, with a post-mortem linked on 1 September. The second was confined to one region, so 5 instead of 0. Rate limits are dynamic, the docs show a 60 requests and 400,000 tokens a minute baseline as an example and send readers to the console for their own, and no fine-tuning job limit was found (8). 429 responses carry `Retry-After` and `x-ratelimit-*` headers, and the guide says to poll at 15 seconds or slower and to recreate a job after a 5xx, with no idempotency key (10). The published SLA document names no Token Factory service level, and the docs say an SLA comes with the Enterprise tier by contract (0). Supervised fine-tuning carries no beta label, but custom model weights for serving are in beta on request (7)." }, { "key": "performance", "name": "Performance", "weight": 10, "effectiveWeight": 0, "pending": true, "points": 0, "reason": "Pending. Latency is measured per call by our probes, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until the first probe window closes." }, { "key": "schema", "name": "Schema \u0026 documentation", "weight": 13, "effectiveWeight": 16.25, "score": 68, "points": 11.05, "reason": "A public OpenAPI 3.1 file at api.tokenfactory.nebius.com/openapi.json covers `/v1/fine_tuning/jobs`, events, checkpoints, files, datasets and operations (25). llms.txt, llms-full.txt and a Markdown twin of every docs page (10). Operation descriptions are one line each. The guide explains each hyperparameter with ranges and when to change it (10). Hyperparameters carry minimums, maximums and defaults, status is a six-value enum and two fields are required, but the request allows additional properties and a free-form `extra_body` (11). Python and cURL examples with sample responses. The reference documents only a 422 validation response, and the guide disagrees with the spec on integrations (it lists only `wandb` as supported and omits the required `api_key`) (7). Paths are versioned `/v1` and the spec is date-stamped 20260930, with no public changelog found (5)." }, { "key": "ergonomics", "name": "Agent ergonomics", "weight": 13, "effectiveWeight": 16.25, "score": 51, "points": 8.29, "reason": "List calls take `limit`, and job objects are compact, with no field selection (12). `limit` and `after` cursors on jobs, events and checkpoints, and a `purpose` filter on files. No filter by status or model found (12). A failed job returns an `error` object with `code`, `message` and `param`, and validation errors name the field. No list of error codes found (10). No idempotency key on job creation. The guide says to recreate a job after a 5xx, and a cancel call exists (5). Only `model` and `training_file` are required and every hyperparameter has a default. There is no Nebius SDK for this API, and the docs use OpenAI's Python and JavaScript clients (12)." }, { "key": "security", "name": "Security \u0026 auth", "weight": 14, "effectiveWeight": 17.5, "score": 52, "points": 9.1, "reason": "Bearer API keys created and deleted per project in the console, shown once. No scopes, expiry or rotation API found in the reviewed documentation (20). Project Admins and Members both have full access to the Files and Fine-tuning APIs, the Billing Manager group has no resource access, and there is no read-only role or confirmation step for deletes (5). Responses carry job state and the customer's own files, no third-party content (10). A usage view by project, service and product. No audit log for Token Factory API calls found (3). security.txt on nebius.com is valid until 31 December 2027, and the trust centre lists SOC 2 Type II, ISO 27001, 27701 and 42001 among others without naming Token Factory in scope. No bug bounty found (14). Hugging Face and Weights \u0026 Biases tokens travel in the job request body and are marked write-only in the spec." }, { "key": "payments", "name": "Payments \u0026 pricing", "weight": 10, "effectiveWeight": 12.5, "score": 0, "points": 0, "reason": "No machine payment protocol (0). No fine-tuning price was found in the docs or the public catalogue JSON, and the product page points to a console price page that is script-drawn and disallowed by robots.txt, which we did not read. Scored as absent for that reason, not as a finding that prices are hidden (0). A bank card is mandatory at onboarding, so the $1 credit for 30 days is not a card-free trial (0). Sign-up, billing and key creation are browser steps (0)." }, { "key": "tasks", "name": "Task success", "weight": 10, "effectiveWeight": 0, "pending": true, "points": 0, "reason": "Pending. Task success needs the category task suites run through each tool, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until then. A data provider's data-quality score is published on its listing now and becomes half of this category when it's scored." }, { "key": "maintenance", "name": "Maintenance \u0026 community", "weight": 7, "effectiveWeight": 8.75, "score": 61, "points": 5.34, "reason": "The OpenAPI file is stamped 20260930 and the fine-tuning reference pages were last modified on 2 October 2026 per the docs sitemap (30). No changelog exists. Dated evidence in the last 90 days is the August deprecation notice, legal documents of 15 and 23 September and 177 docs pages modified in September and October, which earns half (10). Support is a contact form and a Discord community, and the cookbook repository had 127 commits since 10 July 2026 with merged pull requests. We did not read its issue tracker (8). No SDK of its own. The docs rely on current OpenAI clients (8). The cookbook carries tests and a SECURITY.md that points to a different repository (5)." }, { "key": "transparency", "name": "Transparency \u0026 trust", "weight": 7, "effectiveWeight": 8.75, "score": 79, "points": 6.91, "note": "editorial 70, provenance 88", "reason": "Closed service under the Nebius Services Agreement and Token Factory Supplemental Terms, which name the contracting entity by region. The cookbook is MIT (18). The legal guide, DPA and terms agree that content is not used for training and that fine-tuning data is stored in the EU. Inputs and outputs are kept for speculative decoding unless Zero Data Retention is on, no retention period for datasets or checkpoints was found, and the terms allow deletion of tuned models at three days' notice (22). Dated deprecation notices for June and August 2026 name replacements for serverless models. No written minimum notice period found (10). A Token Factory sub-processor list dated 23 September 2026 gives each entity, role, location and transfer mechanism, with 15 days' notice of changes (20)." } ], "assessment": { "date": "2026-10-08", "basis": "public evidence", "confidence": "medium", "notes": { "ergonomics": "List calls take `limit`, and job objects are compact, with no field selection (12). `limit` and `after` cursors on jobs, events and checkpoints, and a `purpose` filter on files. No filter by status or model found (12). A failed job returns an `error` object with `code`, `message` and `param`, and validation errors name the field. No list of error codes found (10). No idempotency key on job creation. The guide says to recreate a job after a 5xx, and a cancel call exists (5). Only `model` and `training_file` are required and every hyperparameter has a default. There is no Nebius SDK for this API, and the docs use OpenAI's Python and JavaScript clients (12).", "maintenance": "The OpenAPI file is stamped 20260930 and the fine-tuning reference pages were last modified on 2 October 2026 per the docs sitemap (30). No changelog exists. Dated evidence in the last 90 days is the August deprecation notice, legal documents of 15 and 23 September and 177 docs pages modified in September and October, which earns half (10). Support is a contact form and a Discord community, and the cookbook repository had 127 commits since 10 July 2026 with merged pull requests. We did not read its issue tracker (8). No SDK of its own. The docs rely on current OpenAI clients (8). The cookbook carries tests and a SECURITY.md that points to a different repository (5).", "payments": "No machine payment protocol (0). No fine-tuning price was found in the docs or the public catalogue JSON, and the product page points to a console price page that is script-drawn and disallowed by robots.txt, which we did not read. Scored as absent for that reason, not as a finding that prices are hidden (0). A bank card is mandatory at onboarding, so the $1 credit for 30 days is not a card-free trial (0). Sign-up, billing and key creation are browser steps (0).", "reliability": "Scored with the hosted lines. status.nebius.com is a Statuspage for the whole Nebius cloud with a Token Factory component in each of nine regions and none for fine-tuning (15 of 20). From 10 July to 8 October 2026 two major incidents were tagged Token Factory, a network incident of 93 minutes on 27 July and overheating hardware in us-central1 from 19 to 20 August lasting about 21.5 hours, with a post-mortem linked on 1 September. The second was confined to one region, so 5 instead of 0. Rate limits are dynamic, the docs show a 60 requests and 400,000 tokens a minute baseline as an example and send readers to the console for their own, and no fine-tuning job limit was found (8). 429 responses carry `Retry-After` and `x-ratelimit-*` headers, and the guide says to poll at 15 seconds or slower and to recreate a job after a 5xx, with no idempotency key (10). The published SLA document names no Token Factory service level, and the docs say an SLA comes with the Enterprise tier by contract (0). Supervised fine-tuning carries no beta label, but custom model weights for serving are in beta on request (7).", "schema": "A public OpenAPI 3.1 file at api.tokenfactory.nebius.com/openapi.json covers `/v1/fine_tuning/jobs`, events, checkpoints, files, datasets and operations (25). llms.txt, llms-full.txt and a Markdown twin of every docs page (10). Operation descriptions are one line each. The guide explains each hyperparameter with ranges and when to change it (10). Hyperparameters carry minimums, maximums and defaults, status is a six-value enum and two fields are required, but the request allows additional properties and a free-form `extra_body` (11). Python and cURL examples with sample responses. The reference documents only a 422 validation response, and the guide disagrees with the spec on integrations (it lists only `wandb` as supported and omits the required `api_key`) (7). Paths are versioned `/v1` and the spec is date-stamped 20260930, with no public changelog found (5).", "security": "Bearer API keys created and deleted per project in the console, shown once. No scopes, expiry or rotation API found in the reviewed documentation (20). Project Admins and Members both have full access to the Files and Fine-tuning APIs, the Billing Manager group has no resource access, and there is no read-only role or confirmation step for deletes (5). Responses carry job state and the customer's own files, no third-party content (10). A usage view by project, service and product. No audit log for Token Factory API calls found (3). security.txt on nebius.com is valid until 31 December 2027, and the trust centre lists SOC 2 Type II, ISO 27001, 27701 and 42001 among others without naming Token Factory in scope. No bug bounty found (14). Hugging Face and Weights \u0026 Biases tokens travel in the job request body and are marked write-only in the spec.", "transparency": "Closed service under the Nebius Services Agreement and Token Factory Supplemental Terms, which name the contracting entity by region. The cookbook is MIT (18). The legal guide, DPA and terms agree that content is not used for training and that fine-tuning data is stored in the EU. Inputs and outputs are kept for speculative decoding unless Zero Data Retention is on, no retention period for datasets or checkpoints was found, and the terms allow deletion of tuned models at three days' notice (22). Dated deprecation notices for June and August 2026 name replacements for serverless models. No written minimum notice period found (10). A Token Factory sub-processor list dated 23 September 2026 gives each entity, role, location and transfer mechanism, with 15 days' notice of changes (20)." }, "sources": [ { "what": "docs index", "url": "https://docs.tokenfactory.nebius.com/llms.txt", "seen": "2026-10-08" }, { "what": "post-training overview", "url": "https://docs.tokenfactory.nebius.com/post-training/overview.md", "seen": "2026-10-08" }, { "what": "models for fine-tuning", "url": "https://docs.tokenfactory.nebius.com/post-training/models.md", "seen": "2026-10-08" }, { "what": "supervised fine-tuning guide", "url": "https://docs.tokenfactory.nebius.com/post-training/how-to-fine-tune.md", "seen": "2026-10-08" }, { "what": "dataset formats", "url": "https://docs.tokenfactory.nebius.com/post-training/datasets.md", "seen": "2026-10-08" }, { "what": "custom model weights (beta)", "url": "https://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/custom-weights.md", "seen": "2026-10-08" }, { "what": "older fine-tuning overview page", "url": "https://docs.tokenfactory.nebius.com/fine-tuning/overview.md", "seen": "2026-10-08" }, { "what": "OpenAPI file", "url": "https://api.tokenfactory.nebius.com/openapi.json", "seen": "2026-10-08" }, { "what": "API introduction and authentication", "url": "https://docs.tokenfactory.nebius.com/api-reference/introduction.md", "seen": "2026-10-08" }, { "what": "rate limits", "url": "https://docs.tokenfactory.nebius.com/ai-models-inference/rate-limits.md", "seen": "2026-10-08" }, { "what": "billing and trial credit", "url": "https://docs.tokenfactory.nebius.com/other-capabilities/billing-new.md", "seen": "2026-10-08" }, { "what": "dedicated endpoint billing policy", "url": "https://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/billing-policy.md", "seen": "2026-10-08" }, { "what": "groups and access", "url": "https://docs.tokenfactory.nebius.com/team-access/groups.md", "seen": "2026-10-08" }, { "what": "legal quick guide", "url": "https://docs.tokenfactory.nebius.com/legal/legal-quick-guide.md", "seen": "2026-10-08" }, { "what": "August 2026 deprecation notice", "url": "https://docs.tokenfactory.nebius.com/august-2026-deprecation-notice.md", "seen": "2026-10-08" }, { "what": "docs sitemap with modification dates", "url": "https://docs.tokenfactory.nebius.com/sitemap.xml", "seen": "2026-10-08" }, { "what": "public model catalogue", "url": "https://tokenfactory.nebius.com/api/public/models_info", "seen": "2026-10-08" }, { "what": "robots.txt of the console host", "url": "https://tokenfactory.nebius.com/robots.txt", "seen": "2026-10-08" }, { "what": "product page", "url": "https://nebius.com/services/token-factory/fine-tuning", "seen": "2026-10-08" }, { "what": "Nebius Services Agreement", "url": "https://docs.nebius.com/legal/agreement", "seen": "2026-10-08" }, { "what": "Token Factory Supplemental Terms", "url": "https://docs.nebius.com/legal/token-factory", "seen": "2026-10-08" }, { "what": "service level agreement", "url": "https://docs.nebius.com/legal/sla", "seen": "2026-10-08" }, { "what": "privacy policy", "url": "https://docs.nebius.com/legal/privacy", "seen": "2026-10-08" }, { "what": "Token Factory sub-processors", "url": "https://docs.nebius.com/legal/sub-processors_tofa", "seen": "2026-10-08" }, { "what": "status page incidents", "url": "https://status.nebius.com/api/v2/incidents.json", "seen": "2026-10-08" }, { "what": "status page components", "url": "https://status.nebius.com/api/v2/components.json", "seen": "2026-10-08" }, { "what": "security.txt", "url": "https://nebius.com/.well-known/security.txt", "seen": "2026-10-08" }, { "what": "trust centre", "url": "https://nebius.com/trust-center", "seen": "2026-10-08" }, { "what": "cookbook repository", "url": "https://github.com/nebius/token-factory-cookbook", "seen": "2026-10-08" }, { "what": "domain registration", "url": "https://rdap.verisign.com/com/v1/domain/nebius.com", "seen": "2026-10-08" } ], "openQuestions": [ "unchecked: the fine-tuning price list. The product page points to https://tokenfactory.nebius.com/organization/prices, which is drawn by script inside the console and disallowed by robots.txt, so Payments scores pricing as absent on our inability to read it.", "unchecked: GitHub stars and the issue tracker of nebius/token-factory-cookbook. The GitHub API refused the one request we made.", "Whether serverless LoRA deployment with per-token billing still exists. An older docs page and the product page describe it, the current models page says dedicated endpoints only, and the deployment guide returns 404.", "Whether the trust centre's SOC 2 and ISO certificates cover Token Factory. The legal guide says so in general terms and the trust centre page does not name the product.", "Whether failed or cancelled fine-tuning jobs are billed, and how many jobs can run at once, were not found in the reviewed documentation.", "The trust centre and product page were read through a summarising fetch. The product page wording on deployment was confirmed against the raw HTML.", "The lead's docs link (`fine-tuning/overview`) still loads but is an older page outside the docs index. The current section is `post-training/overview`." ] }, "negative": -2, "negativeNotes": [ "-2: on 2026-10-08 the product page said a fine-tuned model goes live with one click on 'serverless endpoints, on-demand GPU, or dedicated enterprise clusters', while the docs say deployment is by dedicated endpoints only and custom model weights are in beta on request, and the guide's deployment links return 404 (https://nebius.com/services/token-factory/fine-tuning, https://docs.tokenfactory.nebius.com/post-training/models.md)" ], "verdict": "Supervised fine-tuning on 49 open base models through OpenAI-style `/v1/fine_tuning/jobs` calls, with LoRA or full weights and every checkpoint file downloadable. No fine-tuning price was found outside the script-drawn console, and the docs say tuned models deploy only to dedicated endpoints, with custom weights in beta on request.", "bestFor": "Teams that want supervised LoRA or full fine-tuning of a wide list of open models, up to Qwen3 Coder 480B and DeepSeek, through OpenAI-style calls, with EU storage and the weights to take away.", "strengths": [ "49 base models listed, 42 with LoRA and full fine-tuning and 7 with full fine-tuning only, at context lengths from 8,192 to 131,072 tokens", "Checkpoint files download through `GET /v1/files/{file_id}/content`, and an `hf` integration pushes the result to a Hugging Face repository", "A public OpenAPI 3.1 file covers the fine-tuning, files, datasets and operations paths, with ranges on every hyperparameter", "The legal guide says content is not used to train models and that customers own the models they fine-tune", "A dated sub-processor list for Token Factory, with 15 days' notice of changes, and a valid security.txt on nebius.com" ], "weaknesses": [ "No fine-tuning price found in the docs or the public catalogue JSON. The price page is a script-drawn console page that robots.txt disallows", "The models page says deployment is by dedicated endpoints only, and custom model weights are in beta and available on request", "A bank card is mandatory at onboarding, so the $1 trial credit (30 days) is not a card-free trial", "Two major incidents tagged Token Factory in 90 days, 93 minutes on 27 July and about 21.5 hours in us-central1 from 19 August 2026", "No public changelog, no idempotency key on job creation and only a 422 response documented in the reference", "The Services Agreement (clause 4.1.10) forbids competitive analysis or benchmarking" ], "agentNotes": [ "Use the OpenAI client with `base_url` `https://api.tokenfactory.nebius.com/v1/` and `NEBIUS_API_KEY`. Upload JSONL with `purpose=fine-tune`, then create the job.", "Set `hyperparameters.lora` to true for an adapter. The default is false, which runs full fine-tuning.", "Poll `GET /v1/fine_tuning/jobs/{job_id}` no faster than every 15 seconds. There is no idempotency key, so list jobs before recreating one after a timeout.", "Download every ID in a checkpoint's `result_files` before relying on hosted copies. The terms allow deletion of tuned models at three days' notice.", "The spec requires `wandb.api_key` although the guide omits it, and it also accepts `mlflow` and `hf` integrations. Check the price in the console before starting a job." ], "metrics": { "kind": "remote", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "medium", "grade": "D", "methodology": "0.4", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 47.7 } ], "editorialScores": { "ergonomics": 51, "maintenance": 61, "payments": 0, "reliability": 45, "schema": 68, "security": 52, "transparency": 70 }, "provenanceScore": 88 }, "connect": { "install": "pip3 install --upgrade openai", "http": "curl 'https://api.tokenfactory.nebius.com/v1/fine_tuning/jobs' \\\n -X POST \\\n -H 'Accept: application/json' \\\n -H 'Content-Type: application/json' \\\n -H \"Authorization: Bearer $NEBIUS_API_KEY\" \\\n -d '{\"model\":\"meta-llama/Llama-3.1-8B-Instruct\",\"suffix\":\"my-domain-adapter\",\"training_file\":\"\u003ctraining_file_ID\u003e\",\"hyperparameters\":{\"n_epochs\":3,\"lora\":true,\"lora_r\":16,\"lora_alpha\":16}}'" }, "letme": { "capability": "https://letme.dev/finetune.sft", "tool": "https://letme.dev/nebius-token-factory-fine-tuning" }, "sameCompany": [ "nebius-ai-cloud" ], "notable": [ "The models page lists 49 base models (DeepSeek V3 and V4 Flash, Gemma 4, gpt-oss 20B and 120B, Qwen2.5, Qwen3, Qwen3.5 and Qwen3.6, Llama 3.1 to 3.3) and ends 'Deployment options currently only include via Dedicated endpoints' (https://docs.tokenfactory.nebius.com/post-training/models.md)", "Custom model weights are 'currently in beta and available on request' through the support team (https://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/custom-weights.md)", "The fine-tuning guide links to `post-training/deploy-custom-model` and `fine-tuning/deploy-custom-model` for deployment, and both returned 404 on 8 October 2026. An older page at `fine-tuning/overview`, absent from llms.txt, still describes serverless LoRA deployment with per-token billing (https://docs.tokenfactory.nebius.com/post-training/how-to-fine-tune.md)", "The post-training overview lists supervised fine-tuning and custom speculator training, and names reinforcement fine-tuning as an upcoming limited professional service on request (https://docs.tokenfactory.nebius.com/post-training/overview.md)", "Fine-tuning datasets, artefacts and model outputs are stored only in EU data centres. Jobs from EU users run in the EU and jobs from US users in the US, with no region choice (https://docs.tokenfactory.nebius.com/legal/legal-quick-guide.md)", "The Token Factory Supplemental Terms (effective 15 September 2026) let Nebius delete a customer's fine-tuned model at three days' written notice and cap how many can be in use (https://docs.nebius.com/legal/token-factory)", "Clause 4.1.10 of the Nebius Services Agreement (effective 28 September 2026) forbids use of the services to 'engage in competitive analysis or benchmarking' (https://docs.nebius.com/legal/agreement)", "Dataset files are JSONL in conversational, instruction, text or pre-tokenised form, up to 20GB through the Files API and 50GB through Data Lab (https://docs.tokenfactory.nebius.com/post-training/datasets.md)" ], "area": "models", "details": [ { "label": "Methods", "value": "Supervised fine-tuning, LoRA or full weights, and custom speculator (`spec-draft`) training. Jobs can continue from a checkpoint or a Hugging Face repository (`from_checkpoint`). Reinforcement fine-tuning is a professional service on request" }, { "label": "Base models", "value": "49 listed. 42 take LoRA and full fine-tuning, 7 (DeepSeek, Gemma 4, Qwen3.5 27B, Qwen3.6 27B) full fine-tuning only" }, { "label": "Context length", "value": "8,192 by default, up to 131,072 tokens" }, { "label": "Weights", "value": "Yes. Each checkpoint lists `result_files` to download through the Files API, and an `hf` integration exports to a Hugging Face repository" }, { "label": "Serving", "value": "Dedicated endpoints only, per the models page. Custom model weights are in beta and available on request" }, { "label": "Data", "value": "JSONL in conversational, instruction, text or pre-tokenised form. 20GB through the Files API, 50GB through Data Lab. Stored in the EU" }, { "label": "Hyperparameters", "value": "`n_epochs` 1 to 20, `batch_size` 1 to 64, `lora_r` 8 to 128, plus learning rate, warm-up, weight decay, packing and gradient clipping" }, { "label": "Free tier", "value": "$1 trial credit for 30 days, after a bank card is added" } ], "provenance": { "legalEntity": "Nebius B.V.", "domain": "nebius.com", "domainRegistered": "2004-06-26", "endpointOnVendorDomain": true, "terms": "https://docs.nebius.com/legal/agreement", "privacy": "https://docs.nebius.com/legal/privacy", "statusPage": "https://status.nebius.com", "changelog": "", "securityTxt": "valid", "checked": "2026-10-08", "notes": [ "The Nebius Services Agreement (published 15 September 2026, effective 28 September 2026) names Nebius B.V. as the contracting entity by default, Nebius Inc. for US customers who registered from 15 September 2026 and Nebius Israel Ltd for some customers in Israel. The parent is Nebius Group N.V.", "The Token Factory Supplemental Terms at https://docs.nebius.com/legal/token-factory are incorporated into the agreement and carry the fine-tuning clauses.", "The privacy policy (effective 23 September 2026) names Nebius Token Factory in its scope and covers data Nebius holds as controller. Customer content is covered by the DPA at https://docs.nebius.com/legal/dpa.", "security.txt at nebius.com gives security@nebius.com and expires on 31 December 2027. The same path on tokenfactory.nebius.com returns the console's HTML shell.", "status.nebius.com is an Atlassian Statuspage for the whole Nebius cloud, with a Token Factory component in each of nine regions and no separate fine-tuning component.", "No public changelog for Token Factory was found in the docs index. The OpenAPI file carries the version stamp 20260930-cfb76be12.", "nebius.com was registered on 26 June 2004 per Verisign RDAP." ], "score": 88, "checks": [ { "check": "Legal entity named", "value": "Nebius B.V.", "points": 20, "max": 20, "state": "ok" }, { "check": "Domain age", "value": "nebius.com, registered 2004-06-26 (22 years)", "points": 15, "max": 15, "state": "ok" }, { "check": "Endpoint on the vendor's domain", "value": "api.tokenfactory.nebius.com", "points": 15, "max": 15, "state": "ok" }, { "check": "Terms of service", "value": "read, states 7 of the 7 things a reader expects, and has 1 clause that costs points", "points": 8, "max": 10, "state": "part" }, { "check": "Privacy policy", "value": "read, states 8 of the 8 things a reader expects", "points": 10, "max": 10, "state": "ok" }, { "check": "Status page", "value": "status.nebius.com", "points": 10, "max": 10, "state": "ok" }, { "check": "Changelog", "value": "not found", "points": 0, "max": 10, "state": "no" }, { "check": "security.txt", "value": "valid", "points": 10, "max": 10, "state": "ok" } ], "policies": [ { "kind": "terms", "url": "https://docs.nebius.com/legal/agreement", "state": "read", "readAt": "2026-10-08", "statedDate": "2026-09-28", "words": 12310, "points": 8, "max": 10, "expected": [ { "key": "terms.date", "label": "Gives the date it was last updated", "found": true, "quote": "Effective date: September 28, 2026", "says": "Last updated 2026-09-28" }, { "key": "terms.law", "label": "Names the governing law or courts", "found": true, "quote": "(v) this Agreement is governed and construed in accordance with the laws of the State of Israel, without regard to its conflict of laws rules.", "says": "The law of the State of Israel" }, { "key": "terms.liability", "label": "States a limit on its liability", "found": true, "quote": "…MAXIMUM EXTENT PERMITTED UNDER APPLICABLE LAW, EXCEPT FOR GROSS NEGLIGENCE OR INTENTIONAL MISCONDUCT, IN NO EVENT WILL NEBIUS’ CUMULATIVE AGGREGATE LIABILITY TO CUSTOMER EXCEED 100% OF THE FEES PAID BY CUSTOMER TO NEBIUS FOR THE SERVICES DURING THE 12-MONTH PERIOD IMMEDIATELY PRECEDING THE OCCURRENCE OF THE CLAIM GIVI…", "says": "Capped at the fees paid in the 12 months before the claim" }, { "key": "terms.termination", "label": "Says how the agreement or account can be ended", "found": true, "quote": "If the Customer does not agree with the changes to the Agreement, Linked Documents or pricing, the Customer may terminate this Agreement by sending a written notice of termination within ten (10) calendar days since the changes become effective." }, { "key": "terms.changes", "label": "Says how changes to the terms are announced", "found": true, "quote": "Nebius will inform the Customer at least ten (10) calendar days prior to any material changes to the terms of the Agreement, or the Linked Documents become effective, and at least three (3) calendar days prior to any increase in Service Rates becomes effective.", "says": "Gives ten calendar days of notice before a change" }, { "key": "terms.use", "label": "Lists what users may not do", "found": true, "quote": "The Customer shall not include personal data or confidential or sensitive information in any of these items." }, { "key": "terms.sla", "label": "Refers to a service level or uptime commitment", "found": true, "quote": "Service Level Agreement (“SLA”) is set forth here: https://docs.nebius.com/legal/sla" } ], "toKnow": [ { "key": "terms.benchmark", "label": "Restricts benchmarking or competitive use", "found": true, "quote": "…or improve a product or service that competes with the Services or engage in competitive analysis or benchmarking, or (d) for any illegal, unlawful, fraudulent, unfair, deceptive or other prohibited purposes (including, but not limited to, providing an illegal service, terrorism, illegal hate speech, child pornography…", "costsPoints": true }, { "key": "terms.cutoff", "label": "Says access can be ended without notice or for any reason", "found": true, "quote": "Nebius may terminate the Agreement for cause with the Services being immediately disabled and with no expenses or damages reimbursed without notice if:" }, { "key": "terms.arbitration", "label": "Requires arbitration or waives class actions", "found": true, "quote": "If the arbitrator(s) determine a party to be the prevailing party under circumstances where the prevailing party won on some but not all of the claims and counterclaims, the arbitrator may award the prevailing party an appropriate percentage of the costs and attorneys’ fees reasonably incurred by the prevailing party…" } ], "notes": [ { "date": "2026-10-08", "text": "Nebius may use the customer's name, logo and trademark in advertising and marketing, including customer lists and case studies, with no further consent.", "quote": "The Customer hereby authorizes Nebius to use the Customer’s name, logo, trademark, trade name, and/or the name of the Customer’s software product or website for informational, advertising, and marketing purposes." }, { "date": "2026-10-08", "text": "Customer data on the platform is marked and deleted within 72 hours after the agreement ends, unless applicable law sets another storage period.", "quote": "In case of termination of the Agreement the Customer Data uploaded on the resources of the Platform is marked and deleted along with resources of the Platform used by the Customer within 72 hours after termination of the Agreement unless applicable law stipulates any other storage period." }, { "date": "2026-10-08", "text": "A third party authorised to manage the services for the customer must accept the agreement, and the customer answers for all activity under its account.", "quote": "If the Customer authorizes any third parties to manage the Services on behalf of the Customer, the Customer shall ensure that such third parties accept this Agreement, including the Linked Documents referred to in the Agreement." } ] }, { "kind": "privacy", "url": "https://docs.nebius.com/legal/privacy", "state": "read", "readAt": "2026-10-08", "statedDate": "2026-09-23", "words": 8096, "points": 10, "max": 10, "expected": [ { "key": "privacy.date", "label": "Gives the date it was last updated", "found": true, "quote": "Effective date: September 23, 2026", "says": "Last updated 2026-09-23" }, { "key": "privacy.collected", "label": "Says what personal data is collected", "found": true, "quote": "…service credentials, access permissions, roles, invitations, or similar developer-access features, we may collect and process data associated with those features, including user identifiers, account and organization identifiers, access settings, permission levels, timestamps, usage status, and related security or audi…" }, { "key": "privacy.retention", "label": "Says how long data is kept", "found": true, "quote": "When we process partially obscured copies of your ID for KYC purposes, we retain them for a maximum of 18 months in order to maintain the necessary records for our yearly audits.", "says": "Names a period of 18 months" }, { "key": "privacy.processors", "label": "Says who else receives the data", "found": true, "quote": "The type or identity of third parties to which we disclose personal information under the Data Privacy Framework, and the purposes for which we do so" }, { "key": "privacy.sale", "label": "Says whether personal data is sold or shared for advertising", "found": true, "quote": "We do not sell your personal information to third parties.", "says": "Says it does not sell personal data" }, { "key": "privacy.rights", "label": "Says what rights people have over their data", "found": true, "quote": "According to the GDPR, you have certain rights as a data subject, including the right to access, right to data portability, right to rectification, right to withdraw consent, right to object, right to erasure, right to restriction of processing, right to lodge a complaint, and right to contact a Data Protection Author…" }, { "key": "privacy.contact", "label": "Gives a privacy contact", "found": true, "quote": "If you have any questions or concerns about your privacy, you may contact us or our data protection officer by writing to us at:", "says": "Names a data protection officer" }, { "key": "privacy.transfers", "label": "Says where data is transferred or stored", "found": true, "quote": "Commitment to be subject to the Data Privacy Framework Principles", "says": "Relies on the Data Privacy Framework" } ] } ] }, "pageJsonUrl": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.json", "live": { "slug": "nebius-token-factory-fine-tuning", "probe": { "target": "https://api.tokenfactory.nebius.com/v1", "method": "get", "lastAt": "2026-10-09T09:45:02.51564291Z", "lastOk": true, "lastStatus": 404, "lastMs": 85, "authRequired": false, "uptime24h": 100, "uptime30d": 100, "p50ms24h": 85, "p95ms24h": 134, "samples24h": 23, "samples30d": 23, "days": [ { "date": "2026-10-09", "probes": 23, "ok": 23 } ] }, "vendorStatus": { "page": "https://status.nebius.com", "indicator": "none", "summary": "All Systems Operational", "checkedAt": "2026-10-09T09:47:18.762197795Z" }, "updatedAt": "2026-10-09T09:47:18.762197795Z" } }, "verify": { "accepts": "a page on nebius.com or one of its subdomains, or the README of github.com/nebius/token-factory-cookbook", "badgeUrl": "https://www.anchorterminal.com/badges/nebius-token-factory-fine-tuning.svg", "body": { "slug": "nebius-token-factory-fine-tuning", "url": "the page with the badge or the link" }, "docs": "https://www.anchorterminal.com/builders/#verify", "effect": "none, it never changes a grade, rank or review", "endpoint": "https://www.anchorterminal.com/api/v1/verify", "listingUrl": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning", "mcpTool": "verify_listing", "recheck": "weekly; two failed checks in a row and it lapses, a later pass restores it", "snippets": { "html": "\u003ca href=\"https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning\"\u003e\u003cimg src=\"https://www.anchorterminal.com/badges/nebius-token-factory-fine-tuning.svg\" alt=\"Nebius Token Factory fine-tuning on Anchor Terminal\" height=\"20\"\u003e\u003c/a\u003e", "markdown": "[![Nebius Token Factory fine-tuning on Anchor Terminal](https://www.anchorterminal.com/badges/nebius-token-factory-fine-tuning.svg)](https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning)", "link": "\u003ca href=\"https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning\"\u003eNebius Token Factory fine-tuning on Anchor Terminal\u003c/a\u003e" } } }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning", "json": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.md", "slim": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.min.md" }, "markdown": "## Overview\n\n**Grade D · 47.7/100 · rank #738 of 842 · #9 in Fine-tuning · not agent-ready · confidence medium**\n\n\nMore from Nebius, listed separately because each is its own product: [Nebius AI Cloud](https://www.anchorterminal.com/tools/nebius-ai-cloud.md) (GPU \u0026 serverless compute).\n\n## Assessment\n\nSupervised fine-tuning on 49 open base models through OpenAI-style `/v1/fine_tuning/jobs` calls, with LoRA or full weights and every checkpoint file downloadable. No fine-tuning price was found outside the script-drawn console, and the docs say tuned models deploy only to dedicated endpoints, with custom weights in beta on request.\n\n## Facts\n\n| Field | Value |\n| --- | --- |\n| Vendor | Nebius (https://nebius.com) |\n| Kind | HTTP API |\n| Category | Fine-tuning (https://www.anchorterminal.com/categories/fine-tuning) |\n| Transport | HTTP |\n| Endpoint | `https://api.tokenfactory.nebius.com/v1` |\n| Auth | API key · `Authorization: Bearer` with an API key created by a person in the console under API keys, shown once and read from `NEBIUS_API_KEY` in the docs. Keys belong to a project. Project Admins and Members both have full access to the Files and Fine-tuning APIs, and no per-key scopes or expiry were found in the reviewed documentation. |\n| Pricing | Pay per use (Pay per use) · Pay as you go, with no monthly fee stated. No fine-tuning price was found in the docs or in the public catalogue at `/api/public/models_info`. The product page sends readers to the Token Factory console's prices page, a script-drawn console page that robots.txt disallows, so we did not read it. A bank card is mandatory at onboarding. New accounts get $1 of trial credit valid for 30 days (https://docs.tokenfactory.nebius.com/other-capabilities/billing-new.md). Dedicated endpoints bill while one or more replicas are ready. |\n| x402 | No · No x402, MPP or L402 in the docs index, the OpenAPI file or the billing page (checked 2026-10-08). |\n| Licence | Proprietary service (cookbook examples MIT) |\n| Source | https://github.com/nebius/token-factory-cookbook |\n| Docs | https://docs.tokenfactory.nebius.com/post-training/overview |\n| llms.txt | https://docs.tokenfactory.nebius.com/llms.txt |\n| Last release | 2026-09-30 |\n| Methods | Supervised fine-tuning, LoRA or full weights, and custom speculator (`spec-draft`) training. Jobs can continue from a checkpoint or a Hugging Face repository (`from_checkpoint`). Reinforcement fine-tuning is a professional service on request |\n| Base models | 49 listed. 42 take LoRA and full fine-tuning, 7 (DeepSeek, Gemma 4, Qwen3.5 27B, Qwen3.6 27B) full fine-tuning only |\n| Context length | 8,192 by default, up to 131,072 tokens |\n| Weights | Yes. Each checkpoint lists `result_files` to download through the Files API, and an `hf` integration exports to a Hugging Face repository |\n| Serving | Dedicated endpoints only, per the models page. Custom model weights are in beta and available on request |\n| Data | JSONL in conversational, instruction, text or pre-tokenised form. 20GB through the Files API, 50GB through Data Lab. Stored in the EU |\n| Hyperparameters | `n_epochs` 1 to 20, `batch_size` 1 to 64, `lora_r` 8 to 128, plus learning rate, warm-up, weight decay, packing and gradient clipping |\n| Free tier | $1 trial credit for 30 days, after a bank card is added |\n| Capabilities | finetune.sft, finetune.lora, finetune.export |\n| Tags | hosted, usage-priced, card-required, open-weights, llms-txt, openapi, async-jobs |\n| JSON | https://www.anchorterminal.com/api/v1/tools/nebius-token-factory-fine-tuning.json |\n\n## Score breakdown (methodology v0.4, October 2026 research run)\n\nAssessed 2026-10-08 from public evidence against the published checklist (https://www.anchorterminal.com/benchmark/#checklist). Confidence: medium. Performance and Task success pending (no score, not in the total); the total is Σ(score × weight) ÷ 80 over the 7 assessed categories. \"This run\" is each category's share of the 100 points.\n\n| Category | Weight | This run | Score (0–100) | Points |\n| --- | --- | --- | --- | --- |\n| Reliability | 16% | 20 | 45 | 9.0 |\n| Performance | 10% | pending | pending | n/a |\n| Schema \u0026 documentation | 13% | 16.2 | 68 | 11.1 |\n| Agent ergonomics | 13% | 16.2 | 51 | 8.3 |\n| Security \u0026 auth | 14% | 17.5 | 52 | 9.1 |\n| Payments \u0026 pricing | 10% | 12.5 | 0 | 0.0 |\n| Task success | 10% | pending | pending | n/a |\n| Maintenance \u0026 community | 7% | 8.8 | 61 | 5.3 |\n| Transparency \u0026 trust (editorial 70, provenance 88) | 7% | 8.8 | 79 | 6.9 |\n| Negative events | up to −15 | up to −15 | -2: on 2026-10-08 the product page said a fine-tuned model goes live with one click on 'serverless endpoints, on-demand GPU, or dedicated enterprise clusters', while the docs say deployment is by dedicated endpoints only and custom model weights are in beta on request, and the guide's deployment links return 404 (https://nebius.com/services/token-factory/fine-tuning, https://docs.tokenfactory.nebius.com/post-training/models.md) | -2 |\n| **Total** | | | | **47.7 → D** |\n\n### Why each score\n\n- Reliability 45: Scored with the hosted lines. status.nebius.com is a Statuspage for the whole Nebius cloud with a Token Factory component in each of nine regions and none for fine-tuning (15 of 20). From 10 July to 8 October 2026 two major incidents were tagged Token Factory, a network incident of 93 minutes on 27 July and overheating hardware in us-central1 from 19 to 20 August lasting about 21.5 hours, with a post-mortem linked on 1 September. The second was confined to one region, so 5 instead of 0. Rate limits are dynamic, the docs show a 60 requests and 400,000 tokens a minute baseline as an example and send readers to the console for their own, and no fine-tuning job limit was found (8). 429 responses carry `Retry-After` and `x-ratelimit-*` headers, and the guide says to poll at 15 seconds or slower and to recreate a job after a 5xx, with no idempotency key (10). The published SLA document names no Token Factory service level, and the docs say an SLA comes with the Enterprise tier by contract (0). Supervised fine-tuning carries no beta label, but custom model weights for serving are in beta on request (7).\n- Performance: Pending. Latency is measured per call by our probes, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until the first probe window closes.\n- Schema \u0026 documentation 68: A public OpenAPI 3.1 file at api.tokenfactory.nebius.com/openapi.json covers `/v1/fine_tuning/jobs`, events, checkpoints, files, datasets and operations (25). llms.txt, llms-full.txt and a Markdown twin of every docs page (10). Operation descriptions are one line each. The guide explains each hyperparameter with ranges and when to change it (10). Hyperparameters carry minimums, maximums and defaults, status is a six-value enum and two fields are required, but the request allows additional properties and a free-form `extra_body` (11). Python and cURL examples with sample responses. The reference documents only a 422 validation response, and the guide disagrees with the spec on integrations (it lists only `wandb` as supported and omits the required `api_key`) (7). Paths are versioned `/v1` and the spec is date-stamped 20260930, with no public changelog found (5).\n- Agent ergonomics 51: List calls take `limit`, and job objects are compact, with no field selection (12). `limit` and `after` cursors on jobs, events and checkpoints, and a `purpose` filter on files. No filter by status or model found (12). A failed job returns an `error` object with `code`, `message` and `param`, and validation errors name the field. No list of error codes found (10). No idempotency key on job creation. The guide says to recreate a job after a 5xx, and a cancel call exists (5). Only `model` and `training_file` are required and every hyperparameter has a default. There is no Nebius SDK for this API, and the docs use OpenAI's Python and JavaScript clients (12).\n- Security \u0026 auth 52: Bearer API keys created and deleted per project in the console, shown once. No scopes, expiry or rotation API found in the reviewed documentation (20). Project Admins and Members both have full access to the Files and Fine-tuning APIs, the Billing Manager group has no resource access, and there is no read-only role or confirmation step for deletes (5). Responses carry job state and the customer's own files, no third-party content (10). A usage view by project, service and product. No audit log for Token Factory API calls found (3). security.txt on nebius.com is valid until 31 December 2027, and the trust centre lists SOC 2 Type II, ISO 27001, 27701 and 42001 among others without naming Token Factory in scope. No bug bounty found (14). Hugging Face and Weights \u0026 Biases tokens travel in the job request body and are marked write-only in the spec.\n- Payments \u0026 pricing 0: No machine payment protocol (0). No fine-tuning price was found in the docs or the public catalogue JSON, and the product page points to a console price page that is script-drawn and disallowed by robots.txt, which we did not read. Scored as absent for that reason, not as a finding that prices are hidden (0). A bank card is mandatory at onboarding, so the $1 credit for 30 days is not a card-free trial (0). Sign-up, billing and key creation are browser steps (0).\n- Task success: Pending. Task success needs the category task suites run through each tool, which haven't run yet, so this run doesn't score it. Its weight is shared across the assessed categories until then. A data provider's data-quality score is published on its listing now and becomes half of this category when it's scored.\n- Maintenance \u0026 community 61: The OpenAPI file is stamped 20260930 and the fine-tuning reference pages were last modified on 2 October 2026 per the docs sitemap (30). No changelog exists. Dated evidence in the last 90 days is the August deprecation notice, legal documents of 15 and 23 September and 177 docs pages modified in September and October, which earns half (10). Support is a contact form and a Discord community, and the cookbook repository had 127 commits since 10 July 2026 with merged pull requests. We did not read its issue tracker (8). No SDK of its own. The docs rely on current OpenAI clients (8). The cookbook carries tests and a SECURITY.md that points to a different repository (5).\n- Transparency \u0026 trust 79: Closed service under the Nebius Services Agreement and Token Factory Supplemental Terms, which name the contracting entity by region. The cookbook is MIT (18). The legal guide, DPA and terms agree that content is not used for training and that fine-tuning data is stored in the EU. Inputs and outputs are kept for speculative decoding unless Zero Data Retention is on, no retention period for datasets or checkpoints was found, and the terms allow deletion of tuned models at three days' notice (22). Dated deprecation notices for June and August 2026 name replacements for serverless models. No written minimum notice period found (10). A Token Factory sub-processor list dated 23 September 2026 gives each entity, role, location and transfer mechanism, with 15 days' notice of changes (20).\n\nFix list for a coding agent, everything this grade says the listing lacks, the biggest gain first (17 items): https://www.anchorterminal.com/fixes/nebius-token-factory-fine-tuning.md (JSON https://www.anchorterminal.com/fixes/nebius-token-factory-fine-tuning.json)\n\n### What we couldn't check\n\n- unchecked: the fine-tuning price list. The product page points to https://tokenfactory.nebius.com/organization/prices, which is drawn by script inside the console and disallowed by robots.txt, so Payments scores pricing as absent on our inability to read it.\n- unchecked: GitHub stars and the issue tracker of nebius/token-factory-cookbook. The GitHub API refused the one request we made.\n- Whether serverless LoRA deployment with per-token billing still exists. An older docs page and the product page describe it, the current models page says dedicated endpoints only, and the deployment guide returns 404.\n- Whether the trust centre's SOC 2 and ISO certificates cover Token Factory. The legal guide says so in general terms and the trust centre page does not name the product.\n- Whether failed or cancelled fine-tuning jobs are billed, and how many jobs can run at once, were not found in the reviewed documentation.\n- The trust centre and product page were read through a summarising fetch. The product page wording on deployment was confirmed against the raw HTML.\n- The lead's docs link (`fine-tuning/overview`) still loads but is an older page outside the docs index. The current section is `post-training/overview`.\n\n### Sources\n\n- docs index: \u003chttps://docs.tokenfactory.nebius.com/llms.txt\u003e (seen 2026-10-08)\n- post-training overview: \u003chttps://docs.tokenfactory.nebius.com/post-training/overview.md\u003e (seen 2026-10-08)\n- models for fine-tuning: \u003chttps://docs.tokenfactory.nebius.com/post-training/models.md\u003e (seen 2026-10-08)\n- supervised fine-tuning guide: \u003chttps://docs.tokenfactory.nebius.com/post-training/how-to-fine-tune.md\u003e (seen 2026-10-08)\n- dataset formats: \u003chttps://docs.tokenfactory.nebius.com/post-training/datasets.md\u003e (seen 2026-10-08)\n- custom model weights (beta): \u003chttps://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/custom-weights.md\u003e (seen 2026-10-08)\n- older fine-tuning overview page: \u003chttps://docs.tokenfactory.nebius.com/fine-tuning/overview.md\u003e (seen 2026-10-08)\n- OpenAPI file: \u003chttps://api.tokenfactory.nebius.com/openapi.json\u003e (seen 2026-10-08)\n- API introduction and authentication: \u003chttps://docs.tokenfactory.nebius.com/api-reference/introduction.md\u003e (seen 2026-10-08)\n- rate limits: \u003chttps://docs.tokenfactory.nebius.com/ai-models-inference/rate-limits.md\u003e (seen 2026-10-08)\n- billing and trial credit: \u003chttps://docs.tokenfactory.nebius.com/other-capabilities/billing-new.md\u003e (seen 2026-10-08)\n- dedicated endpoint billing policy: \u003chttps://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/billing-policy.md\u003e (seen 2026-10-08)\n- groups and access: \u003chttps://docs.tokenfactory.nebius.com/team-access/groups.md\u003e (seen 2026-10-08)\n- legal quick guide: \u003chttps://docs.tokenfactory.nebius.com/legal/legal-quick-guide.md\u003e (seen 2026-10-08)\n- August 2026 deprecation notice: \u003chttps://docs.tokenfactory.nebius.com/august-2026-deprecation-notice.md\u003e (seen 2026-10-08)\n- docs sitemap with modification dates: \u003chttps://docs.tokenfactory.nebius.com/sitemap.xml\u003e (seen 2026-10-08)\n- public model catalogue: \u003chttps://tokenfactory.nebius.com/api/public/models_info\u003e (seen 2026-10-08)\n- robots.txt of the console host: \u003chttps://tokenfactory.nebius.com/robots.txt\u003e (seen 2026-10-08)\n- product page: \u003chttps://nebius.com/services/token-factory/fine-tuning\u003e (seen 2026-10-08)\n- Nebius Services Agreement: \u003chttps://docs.nebius.com/legal/agreement\u003e (seen 2026-10-08)\n- Token Factory Supplemental Terms: \u003chttps://docs.nebius.com/legal/token-factory\u003e (seen 2026-10-08)\n- service level agreement: \u003chttps://docs.nebius.com/legal/sla\u003e (seen 2026-10-08)\n- privacy policy: \u003chttps://docs.nebius.com/legal/privacy\u003e (seen 2026-10-08)\n- Token Factory sub-processors: \u003chttps://docs.nebius.com/legal/sub-processors_tofa\u003e (seen 2026-10-08)\n- status page incidents: \u003chttps://status.nebius.com/api/v2/incidents.json\u003e (seen 2026-10-08)\n- status page components: \u003chttps://status.nebius.com/api/v2/components.json\u003e (seen 2026-10-08)\n- security.txt: \u003chttps://nebius.com/.well-known/security.txt\u003e (seen 2026-10-08)\n- trust centre: \u003chttps://nebius.com/trust-center\u003e (seen 2026-10-08)\n- cookbook repository: \u003chttps://github.com/nebius/token-factory-cookbook\u003e (seen 2026-10-08)\n- domain registration: \u003chttps://rdap.verisign.com/com/v1/domain/nebius.com\u003e (seen 2026-10-08)\n\n## Who's behind it (provenance 88/100, checked 2026-10-08)\n\n| Check | Finding | Points |\n| --- | --- | --- |\n| Legal entity named | Nebius B.V. | 20/20 |\n| Domain age | nebius.com, registered 2004-06-26 (22 years) | 15/15 |\n| Endpoint on the vendor's domain | api.tokenfactory.nebius.com | 15/15 |\n| Terms of service | read, states 7 of the 7 things a reader expects, and has 1 clause that costs points | 8/10 |\n| Privacy policy | read, states 8 of the 8 things a reader expects | 10/10 |\n| Status page | status.nebius.com | 10/10 |\n| Changelog | not found | 0/10 |\n| security.txt | valid | 10/10 |\n\nThe Nebius Services Agreement (published 15 September 2026, effective 28 September 2026) names Nebius B.V. as the contracting entity by default, Nebius Inc. for US customers who registered from 15 September 2026 and Nebius Israel Ltd for some customers in Israel. The parent is Nebius Group N.V.\n\nThe Token Factory Supplemental Terms at https://docs.nebius.com/legal/token-factory are incorporated into the agreement and carry the fine-tuning clauses.\n\nThe privacy policy (effective 23 September 2026) names Nebius Token Factory in its scope and covers data Nebius holds as controller. Customer content is covered by the DPA at https://docs.nebius.com/legal/dpa.\n\nsecurity.txt at nebius.com gives security@nebius.com and expires on 31 December 2027. The same path on tokenfactory.nebius.com returns the console's HTML shell.\n\nstatus.nebius.com is an Atlassian Statuspage for the whole Nebius cloud, with a Token Factory component in each of nine regions and no separate fine-tuning component.\n\nNo public changelog for Token Factory was found in the docs index. The OpenAPI file carries the version stamp 20260930-cfb76be12.\n\nnebius.com was registered on 26 June 2004 per Verisign RDAP.\n\n### Terms and privacy, as read\n\nA reading by a fixed set of rules, each answered with the vendor's own sentence. Not legal advice.\n\n**Terms of service** (https://docs.nebius.com/legal/agreement), read 2026-10-08, dated 2026-09-28, states 7 of the 7 things a reader expects.\n\n- To know. Restricts benchmarking or competitive use (costs points). \"…or improve a product or service that competes with the Services or engage in competitive analysis or benchmarking, or (d) for any illegal, unlawful, fraudulent, unfair, deceptive or other prohibited purposes (including, but not limited to, providing an illegal service, terrorism, illegal hate speech, child pornography…\"\n- To know. Says access can be ended without notice or for any reason. \"Nebius may terminate the Agreement for cause with the Services being immediately disabled and with no expenses or damages reimbursed without notice if:\"\n- To know. Requires arbitration or waives class actions. \"If the arbitrator(s) determine a party to be the prevailing party under circumstances where the prevailing party won on some but not all of the claims and counterclaims, the arbitrator may award the prevailing party an appropriate percentage of the costs and attorneys’ fees reasonably incurred by the prevailing party…\"\n- Gives the date it was last updated. Last updated 2026-09-28.\n- Names the governing law or courts. The law of the State of Israel.\n- States a limit on its liability. Capped at the fees paid in the 12 months before the claim.\n- Says how changes to the terms are announced. Gives ten calendar days of notice before a change.\n- Also in the text (2026-10-08). Nebius may use the customer's name, logo and trademark in advertising and marketing, including customer lists and case studies, with no further consent. \"The Customer hereby authorizes Nebius to use the Customer’s name, logo, trademark, trade name, and/or the name of the Customer’s software product or website for informational, advertising, and marketing purposes.\"\n- Also in the text (2026-10-08). Customer data on the platform is marked and deleted within 72 hours after the agreement ends, unless applicable law sets another storage period. \"In case of termination of the Agreement the Customer Data uploaded on the resources of the Platform is marked and deleted along with resources of the Platform used by the Customer within 72 hours after termination of the Agreement unless applicable law stipulates any other storage period.\"\n- Also in the text (2026-10-08). A third party authorised to manage the services for the customer must accept the agreement, and the customer answers for all activity under its account. \"If the Customer authorizes any third parties to manage the Services on behalf of the Customer, the Customer shall ensure that such third parties accept this Agreement, including the Linked Documents referred to in the Agreement.\"\n\n**Privacy policy** (https://docs.nebius.com/legal/privacy), read 2026-10-08, dated 2026-09-23, states 8 of the 8 things a reader expects.\n\n- Gives the date it was last updated. Last updated 2026-09-23.\n- Says how long data is kept. Names a period of 18 months.\n- Says whether personal data is sold or shared for advertising. Says it does not sell personal data.\n- Gives a privacy contact. Names a data protection officer.\n- Says where data is transferred or stored. Relies on the Data Privacy Framework.\n\n## Live (updated 2026-10-09 09:47 UTC)\n\n- Right now: up, HTTP 404, 85 ms, checked 2026-10-09 09:45 UTC (get on `https://api.tokenfactory.nebius.com/v1`)\n- Uptime 24h 100.0% (23 probes) · 30 days 100.0% (23 probes) · p50 85 ms · p95 134 ms\n- Vendor status page: none, All Systems Operational\n- Always current: https://www.anchorterminal.com/api/v1/live/nebius-token-factory-fine-tuning.json\n\n## Probe metrics\n\nNot measured yet. Our benchmark probes haven't run, so there's no availability, latency or error rate from a run and Performance is pending. Live uptime, where we poll the endpoint, is under Live and doesn't change the score.\n\n## Strengths\n\n- 49 base models listed, 42 with LoRA and full fine-tuning and 7 with full fine-tuning only, at context lengths from 8,192 to 131,072 tokens\n- Checkpoint files download through `GET /v1/files/{file_id}/content`, and an `hf` integration pushes the result to a Hugging Face repository\n- A public OpenAPI 3.1 file covers the fine-tuning, files, datasets and operations paths, with ranges on every hyperparameter\n- The legal guide says content is not used to train models and that customers own the models they fine-tune\n- A dated sub-processor list for Token Factory, with 15 days' notice of changes, and a valid security.txt on nebius.com\n\n## Weaknesses\n\n- No fine-tuning price found in the docs or the public catalogue JSON. The price page is a script-drawn console page that robots.txt disallows\n- The models page says deployment is by dedicated endpoints only, and custom model weights are in beta and available on request\n- A bank card is mandatory at onboarding, so the $1 trial credit (30 days) is not a card-free trial\n- Two major incidents tagged Token Factory in 90 days, 93 minutes on 27 July and about 21.5 hours in us-central1 from 19 August 2026\n- No public changelog, no idempotency key on job creation and only a 422 response documented in the reference\n- The Services Agreement (clause 4.1.10) forbids competitive analysis or benchmarking\n\n## Before you call it (notes for agents)\n\n1. Use the OpenAI client with `base_url` `https://api.tokenfactory.nebius.com/v1/` and `NEBIUS_API_KEY`. Upload JSONL with `purpose=fine-tune`, then create the job.\n2. Set `hyperparameters.lora` to true for an adapter. The default is false, which runs full fine-tuning.\n3. Poll `GET /v1/fine_tuning/jobs/{job_id}` no faster than every 15 seconds. There is no idempotency key, so list jobs before recreating one after a timeout.\n4. Download every ID in a checkpoint's `result_files` before relying on hosted copies. The terms allow deletion of tuned models at three days' notice.\n5. The spec requires `wandb.api_key` although the guide omits it, and it also accepts `mlflow` and `hf` integrations. Check the price in the console before starting a job.\n\n## Connect\n\nInstall:\n\n```bash\npip3 install --upgrade openai\n```\n\nFirst request:\n\n```bash\ncurl 'https://api.tokenfactory.nebius.com/v1/fine_tuning/jobs' \\\n -X POST \\\n -H 'Accept: application/json' \\\n -H 'Content-Type: application/json' \\\n -H \"Authorization: Bearer $NEBIUS_API_KEY\" \\\n -d '{\"model\":\"meta-llama/Llama-3.1-8B-Instruct\",\"suffix\":\"my-domain-adapter\",\"training_file\":\"\u003ctraining_file_ID\u003e\",\"hyperparameters\":{\"n_epochs\":3,\"lora\":true,\"lora_r\":16,\"lora_alpha\":16}}'\n```\n\nThrough letme (picks today, calling later): https://letme.dev/nebius-token-factory-fine-tuning. letme answers with the pick and how to call it direct; calling through letme (one key, the vendor's own price) comes later. How it works: https://www.anchorterminal.com/letme/index.md\n\n## Similar tools\n\nRanked by shared capabilities, then score. Same-category tools with no shared capability key are listed last.\n\n| Tool | Grade | Score | Rank | Shared capabilities | x402 | Markdown |\n| --- | --- | --- | --- | --- | --- | --- |\n| Axolotl | B | 64.8 | 307 | finetune.sft, finetune.lora, finetune.export | no | https://www.anchorterminal.com/tools/axolotl.md |\n| Fireworks AI Fine-tuning | C | 59 | 506 | finetune.sft, finetune.lora, finetune.export | no | https://www.anchorterminal.com/tools/fireworks-fine-tuning.md |\n| Together AI Fine-tuning | C | 54.7 | 608 | finetune.sft, finetune.lora, finetune.export | no | https://www.anchorterminal.com/tools/together-fine-tuning.md |\n| Unsloth | D | 51.5 | 668 | finetune.sft, finetune.lora, finetune.export | no | https://www.anchorterminal.com/tools/unsloth.md |\n| Tinker | D | 51 | 677 | finetune.sft, finetune.lora, finetune.export | no | https://www.anchorterminal.com/tools/tinker.md |\n| Vertex AI Gemini tuning | B | 64.2 | 325 | finetune.sft, finetune.lora | no | https://www.anchorterminal.com/tools/vertex-ai-tuning.md |\n\n## Panel reviews (0)\n\nReviewed by the Anchor panel (https://www.anchorterminal.com/reviewers/index.md): .\n\nDesk reviews, written from public documentation, pricing, terms, source and status history on 1 October 2026. No calls made. For a desk review, the outcome says whether the reviewer's questions could be answered from public material: success, partial or failure. How reviews work: https://www.anchorterminal.com/reviews/how-it-works.md\n\n## Notable\n\n- The models page lists 49 base models (DeepSeek V3 and V4 Flash, Gemma 4, gpt-oss 20B and 120B, Qwen2.5, Qwen3, Qwen3.5 and Qwen3.6, Llama 3.1 to 3.3) and ends 'Deployment options currently only include via Dedicated endpoints' (source: \u003chttps://docs.tokenfactory.nebius.com/post-training/models.md\u003e)\n- Custom model weights are 'currently in beta and available on request' through the support team (source: \u003chttps://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/custom-weights.md\u003e)\n- The fine-tuning guide links to `post-training/deploy-custom-model` and `fine-tuning/deploy-custom-model` for deployment, and both returned 404 on 8 October 2026. An older page at `fine-tuning/overview`, absent from llms.txt, still describes serverless LoRA deployment with per-token billing (source: \u003chttps://docs.tokenfactory.nebius.com/post-training/how-to-fine-tune.md\u003e)\n- The post-training overview lists supervised fine-tuning and custom speculator training, and names reinforcement fine-tuning as an upcoming limited professional service on request (source: \u003chttps://docs.tokenfactory.nebius.com/post-training/overview.md\u003e)\n- Fine-tuning datasets, artefacts and model outputs are stored only in EU data centres. Jobs from EU users run in the EU and jobs from US users in the US, with no region choice (source: \u003chttps://docs.tokenfactory.nebius.com/legal/legal-quick-guide.md\u003e)\n- The Token Factory Supplemental Terms (effective 15 September 2026) let Nebius delete a customer's fine-tuned model at three days' written notice and cap how many can be in use (source: \u003chttps://docs.nebius.com/legal/token-factory\u003e)\n- Clause 4.1.10 of the Nebius Services Agreement (effective 28 September 2026) forbids use of the services to 'engage in competitive analysis or benchmarking' (source: \u003chttps://docs.nebius.com/legal/agreement\u003e)\n- Dataset files are JSONL in conversational, instruction, text or pre-tokenised form, up to 20GB through the Files API and 50GB through Data Lab (source: \u003chttps://docs.tokenfactory.nebius.com/post-training/datasets.md\u003e)\n\n## Compare\n\n- [Amazon Bedrock model customisation vs Nebius Token Factory fine-tuning](https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-nebius-token-factory-fine-tuning.md): BB 75.8 vs D 47.7\n- [Axolotl vs Nebius Token Factory fine-tuning](https://www.anchorterminal.com/compare/axolotl-vs-nebius-token-factory-fine-tuning.md): B 64.8 vs D 47.7\n- [Microsoft Foundry fine-tuning (Azure OpenAI) vs Nebius Token Factory fine-tuning](https://www.anchorterminal.com/compare/azure-foundry-fine-tuning-vs-nebius-token-factory-fine-tuning.md): C 61.1 vs D 47.7\n- [Fireworks AI Fine-tuning vs Nebius Token Factory fine-tuning](https://www.anchorterminal.com/compare/fireworks-fine-tuning-vs-nebius-token-factory-fine-tuning.md): C 59 vs D 47.7\n- [Nebius Token Factory fine-tuning vs Tinker](https://www.anchorterminal.com/compare/nebius-token-factory-fine-tuning-vs-tinker.md): D 47.7 vs D 51\n- [Nebius Token Factory fine-tuning vs Together AI Fine-tuning](https://www.anchorterminal.com/compare/nebius-token-factory-fine-tuning-vs-together-fine-tuning.md): D 47.7 vs C 54.7\n- [Nebius Token Factory fine-tuning vs Unsloth](https://www.anchorterminal.com/compare/nebius-token-factory-fine-tuning-vs-unsloth.md): D 47.7 vs D 51.5\n- [Nebius Token Factory fine-tuning vs Vertex AI Gemini tuning](https://www.anchorterminal.com/compare/nebius-token-factory-fine-tuning-vs-vertex-ai-tuning.md): D 47.7 vs B 64.2\n\n## Verify this listing\n\nFor the vendor. The badge or a plain link to this page verifies the listing, from a page on nebius.com or one of its subdomains, or the README of github.com/nebius/token-factory-cookbook. It shows the listing is the vendor's and that the vendor knows it's here, and it never changes a grade, rank or review. The vendor sends the page's address to `POST https://www.anchorterminal.com/api/v1/verify` as `{\"slug\": \"nebius-token-factory-fine-tuning\", \"url\": \"…\"}`, or calls the `verify_listing` tool at https://www.anchorterminal.com/mcp. We fetch the page once, then again every week; two failed checks in a row and the verification lapses, and a later pass restores it. What we check: https://www.anchorterminal.com/builders/index.md#verify\n\nHTML badge:\n\n```html\n\u003ca href=\"https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning\"\u003e\u003cimg src=\"https://www.anchorterminal.com/badges/nebius-token-factory-fine-tuning.svg\" alt=\"Nebius Token Factory fine-tuning on Anchor Terminal\" height=\"20\"\u003e\u003c/a\u003e\n```\n\nMarkdown badge, for a README:\n\n```markdown\n[![Nebius Token Factory fine-tuning on Anchor Terminal](https://www.anchorterminal.com/badges/nebius-token-factory-fine-tuning.svg)](https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning)\n```\n\nPlain link:\n\n```html\n\u003ca href=\"https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning\"\u003eNebius Token Factory fine-tuning on Anchor Terminal\u003c/a\u003e\n```\n\n## Share this listing\n\nFor the vendor. Sharing assets for social media, two PNGs of 1200 × 630 that say Nebius Token Factory fine-tuning is listed on Anchor Terminal, with the vendor's logo and this page's address and no grade or score.\n\n- Dark: https://www.anchorterminal.com/assets/share/nebius-token-factory-fine-tuning-dark.png\n- Light: https://www.anchorterminal.com/assets/share/nebius-token-factory-fine-tuning-light.png\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-10-09", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.4", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Terminal", "url": "https://www.anchorterminal.com/tools/" }, { "name": "Fine-tuning", "url": "https://www.anchorterminal.com/categories/fine-tuning" }, { "name": "Nebius Token Factory fine-tuning", "url": "" } ], "description": "Nebius Token Factory runs supervised fine-tuning jobs on open models such as Llama, Qwen, gpt-oss, Gemma and DeepSeek through an OpenAI-compatible REST API, with LoRA or full weights and downloadable checkpoints.", "facts": [ "rank #738 of 842", "API key auth", "0 desk reviews" ], "h1": "Nebius Token Factory fine-tuning", "image": "https://www.anchorterminal.com/assets/og/tools-nebius-token-factory-fine-tuning.png", "path": "/tools/nebius-token-factory-fine-tuning", "published": "2026-10-01", "section": "tools", "title": "Nebius Token Factory fine-tuning review, grade D (47.7/100)", "toc": null, "updated": "2026-10-09", "url": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning" }, "tokens": { "markdown": 8050, "slim": 1580 }, "version": 1 }