{ "data": { "category": { "area": "models", "capabilities": [ "finetune.sft", "finetune.preference", "finetune.rl", "finetune.lora", "finetune.export" ], "description": "Services that train a model on your examples, by supervised, preference or reinforcement fine-tuning, and serve the result. Compared on which base models you can tune, the methods, price per training token, whether you get the weights and what serving the result costs.", "json": "https://www.anchorterminal.com/categories/fine-tuning.json", "name": "Fine-tuning", "slug": "fine-tuning", "test": "The same small dataset used to tune a comparable open model on each service, then served. We check the job flow, how long training takes, whether the weights can leave, and the training and serving cost.", "title": "Fine-tuning services for AI models", "toolCount": 9, "tools": [ "amazon-bedrock-customization", "axolotl", "vertex-ai-tuning", "azure-foundry-fine-tuning", "fireworks-fine-tuning", "together-fine-tuning", "unsloth", "tinker", "nebius-token-factory-fine-tuning" ], "url": "https://www.anchorterminal.com/categories/fine-tuning" }, "faq": [ { "answer": "Amazon Bedrock model customisation has the highest benchmark score of the 9 ranked fine-tuning services for AI models, 75.8 (BB). Axolotl is second with 64.8 (B).", "question": "What are the highest-rated fine-tuning services for AI models for AI agents?" }, { "answer": "1 of the 9 ranked here grade BB or better, the bar for agent-ready on the Anchor benchmark.", "question": "How many fine-tuning services for AI models are agent-ready?" }, { "answer": "None of the ranked listings here accepts x402 for its main call yet.", "question": "Which fine-tuning services for AI models accept x402 payments?" }, { "answer": "By the Anchor benchmark score out of 100, a weighted mean of the scored categories minus deductions for negative events, from public evidence re-checked as vendors change. Listings cannot pay for a place. The latest assessment behind this page is from 8 October 2026.", "question": "How is this list ranked?" } ], "howToChoose": [ { "label": "Weight export and licence", "detail": "Check whether the tuned weights can be downloaded and under what licence, because an agent that must run offline or on your own hardware needs them." }, { "label": "Tunable base models and methods", "detail": "Check which base models can be tuned and whether your chosen method is supported for each model, since method support can differ between them." }, { "label": "Training price by method", "detail": "Check the training price per token for supervised and reinforcement runs separately, because the reinforcement price can differ from the supervised price on the same data." }, { "label": "Serving and idle costs", "detail": "Check the serving price, idle deletion and checkpoint storage fees, because an agent that calls a tuned model only occasionally can end up paying for idle time." } ], "picks": [ { "also": { "name": "Axolotl", "slug": "axolotl", "why": "B, 64.8/100" }, "name": "Amazon Bedrock model customisation", "need": "Highest score overall", "slug": "amazon-bedrock-customization", "why": "BB, 75.8/100 on the benchmark" }, { "name": "Axolotl", "need": "Maintenance \u0026 community", "slug": "axolotl", "why": "88/100 on maintenance \u0026 community, against 45 for the overall leader" }, { "name": "Vertex AI Gemini tuning", "need": "Transparency \u0026 trust", "slug": "vertex-ai-tuning", "why": "88/100 on transparency \u0026 trust, against 83 for the overall leader" }, { "also": { "name": "Unsloth", "slug": "unsloth", "why": "self-hosted, Apache-2 licence" }, "name": "Axolotl", "need": "Self-hosting under an open licence", "slug": "axolotl", "why": "self-hosted, Apache-2 licence" } ], "ranked": 9, "shortlist": [ { "bestFor": "Teams already on AWS that want to tune Amazon Nova or Llama models, or run reinforcement fine-tuning with a Lambda reward function, and serve the result inside Bedrock.", "grade": "BB", "name": "Amazon Bedrock model customisation", "position": 1, "price": "Pay per use", "score": 75.8, "slug": "amazon-bedrock-customization", "strengths": [ "`CreateModelCustomizationJob` accepts a `clientRequestToken`, so a repeated create after a timeout doesn't start a second job", "`ListModelCustomizationJobs` filters by status, name and creation time, sorts, and pages with `maxResults` up to 1,000 and `nextToken`", "Service Terms section 50.12.4 gives the customer exclusive use of a customised model and bars third-party model providers from accessing it" ], "url": "https://www.anchorterminal.com/tools/amazon-bedrock-customization", "verdict": "Job creation takes an idempotency token, job lists filter and paginate, and the Service Terms give the customer exclusive use of a tuned model. Jobs run in two US Regions only, weights can't be exported, and the newest customisation API change found dates from 28 May 2026.", "weaknesses": [ "Fine-tuning runs in us-east-1 and us-west-2 only, and each base model in one of them (two for the Titan models)", "No weight export was found in the reviewed documentation, and Service Terms section 50.11 forbids extracting model weights", "Setup needs an IAM service role, `iam:PassRole` and S3 buckets for input and output before the first job" ], "where": "hosted", "x402": "no" }, { "bestFor": "A team that wants a repeatable, config-driven fine-tune of an open model on its own or rented GPUs, including multi-GPU and multi-node runs.", "grade": "B", "name": "Axolotl", "position": 2, "price": "Free · OSS", "score": 64.8, "slug": "axolotl", "strengths": [ "Apache-2.0, free, and the weights stay on the owner's hardware", "`axolotl config-schema` prints the full config as JSON Schema, and `axolotl agent-docs` prints bundled Markdown references by topic", "SFT, LoRA, QLoRA, DPO, IPO, KTO, ORPO, GRPO and reward modelling from one config format" ], "url": "https://www.anchorterminal.com/tools/axolotl", "verdict": "Axolotl runs a whole fine-tuning job from one YAML file and ships a JSON Schema of its config plus bundled agent docs. It is 0.x software with telemetry on by default, no terms or privacy policy, and the owner supplies the GPU.", "weaknesses": [ "Telemetry to PostHog is on by default and delays training start by 10 seconds until the variable is set either way", "No terms of service, privacy policy, legal entity or security.txt found on axolotl.ai", "Version 0.20.0, with removals in minor releases (FSDP1 in 0.20.0, `relora_steps` renamed in 0.17.0 with no shim)" ], "where": "library", "x402": "no" }, { "bestFor": "Teams already on Google Cloud who need to tune Gemini itself, especially with RL, and will serve it there.", "grade": "B", "name": "Vertex AI Gemini tuning", "position": 3, "price": "Pay per use", "score": 64.2, "slug": "vertex-ai-tuning", "strengths": [ "Supervised, preference and reinforcement tuning of Gemini, plus supervised tuning of Gemma, Llama and Qwen", "No Vertex or Gemini incidents on the Google Cloud status dashboard from July to September 2026", "Public proto for GenAiTuningService with filter and pagination on job lists, and docs pages served as Markdown at `.md.txt`" ], "url": "https://www.anchorterminal.com/tools/vertex-ai-tuning", "verdict": "Supervised, preference and reinforcement tuning of Gemini, plus supervised tuning of Gemma, Llama and Qwen. No weight export. The tuned model exists only as a Google Cloud endpoint.", "weaknesses": [ "No weight export. The tuned model exists only as a Google Cloud endpoint", "Tuned Gemini 3 inference costs 1.5x the base model for as long as you serve it", "Setup needs a project, billing, IAM and a Cloud Storage bucket before the first job" ], "where": "hosted", "x402": "no" }, { "bestFor": "Teams that must tune an OpenAI model, need Azure's compliance and regional controls, and will serve the result on Azure.", "grade": "C", "name": "Microsoft Foundry fine-tuning (Azure OpenAI)", "position": 4, "price": "Pay per use", "score": 61.1, "slug": "azure-foundry-fine-tuning", "strengths": [ "SFT, DPO and RFT on GPT-4.1 and o4-mini through the OpenAI-shaped /openai/v1 API", "Retirement policy with 60 days' notice and published training and deployment retirement dates per tunable model", "Entra ID with RBAC, Azure Monitor logs and an activity log for every customer" ], "url": "https://www.anchorterminal.com/tools/azure-foundry-fine-tuning", "verdict": "SFT, DPO and RFT on GPT-4.1 and o4-mini through the OpenAI-shaped /openai/v1 API. No weight export; checkpoints copy only between Azure resources.", "weaknesses": [ "No weight export; checkpoints copy only between Azure resources", "$1.70 an hour hosting on Standard deployments, and deletion after 15 idle days", "GPT-4.1 training at $25 per 1M tokens globally, and no free tier without a card" ], "where": "hosted", "x402": "no" }, { "bestFor": "Teams that want managed SFT, DPO or RFT on large open models and may later write a custom RL loop on the same platform.", "grade": "C", "name": "Fireworks AI Fine-tuning", "position": 5, "price": "Pay per use", "score": 59, "slug": "fireworks-fine-tuning", "strengths": [ "SFT, DPO, ORPO and RFT as managed jobs, plus a serverless Training API that is generally available", "LoRA SFT from $0.50 per 1M training tokens up to 16B parameters, with serving at base-model prices", "List endpoints take readMask, pageSize up to 200, AIP-160 filters and orderBy" ], "url": "https://www.anchorterminal.com/tools/fireworks-fine-tuning", "verdict": "SFT, DPO, ORPO and RFT as managed jobs, plus a serverless Training API that is generally available. Tuned LoRAs only deploy to on-demand GPUs at $8 an hour and up, never to serverless.", "weaknesses": [ "Tuned LoRAs only deploy to on-demand GPUs at $8 an hour and up, never to serverless", "No training without a payment method; the $1 sign-up credit buys inference only", "The status page has no training component; a 46-hour cloud provider incident in August hit dedicated deployments" ], "where": "hosted", "x402": "no" }, { "bestFor": "Teams that want to tune a large open model, possibly with full fine-tuning, and take the weights away.", "grade": "C", "name": "Together AI Fine-tuning", "position": 6, "price": "Pay per use", "score": 54.7, "slug": "together-fine-tuning", "strengths": [ "31 tunable base models, 11 or 12 of them with full fine-tuning as well as LoRA", "GET /v1/finetune/download returns merged weights or the adapter, at any saved checkpoint", "POST /v1/fine-tunes/estimate-price quotes a job before it runs" ], "url": "https://www.anchorterminal.com/tools/together-fine-tuning", "verdict": "31 tunable base models, 11 or 12 of them with full fine-tuning as well as LoRA. Fine-tuned models don't run serverless; dedicated endpoints start at $5.49 an hour.", "weaknesses": [ "Fine-tuned models don't run serverless; dedicated endpoints start at $5.49 an hour", "No free trial, a $5 prepaid purchase before the first call, and job minimums up to $60", "The status page covers serverless models only, and no fine-tuning rate limits are published" ], "where": "hosted", "x402": "no" }, { "bestFor": "One person or a small team tuning an open model on their own GPU and keeping the weights.", "grade": "D", "name": "Unsloth", "position": 7, "price": "Free · OSS", "score": 51.5, "slug": "unsloth", "strengths": [ "Free and open source, Apache-2.0 core, with the weights staying on your hardware", "LoRA, QLoRA, full fine-tuning, GRPO, DPO and ORPO from one package", "Exports adapters, merged 16-bit weights and GGUF for vLLM, Ollama or llama.cpp" ], "url": "https://www.anchorterminal.com/tools/unsloth", "verdict": "The Apache-2.0 core runs on customer hardware and keeps model weights there. Users supply and pay for the GPU.", "weaknesses": [ "Not a hosted service; you bring and pay for the GPU", "Studio is AGPL-3.0, and its server-side tools are on by default when exposed", "792 open issues and 472 open pull requests" ], "where": "library", "x402": "no" }, { "bestFor": "Researchers and teams writing custom post-training loops, especially RL, who want per-token billing and the weights at the end.", "grade": "D", "name": "Tinker", "position": 8, "price": "Pay per use", "score": 51, "slug": "tinker", "strengths": [ "Full control of the training loop with the GPUs abstracted away, plus recipes for SFT, DPO, RL and distillation", "Checkpoints download and merge into Hugging Face safetensors, so the weights can leave", "Per-token billing with machine-readable prices in models.json" ], "url": "https://www.anchorterminal.com/tools/tinker", "verdict": "Full control of the training loop with the GPUs abstracted away, plus recipes for SFT, DPO, RL and distillation. LoRA only; no full-parameter training.", "weaknesses": [ "LoRA only; no full-parameter training", "Python SDK only, with no REST reference or OpenAPI", "No terms of service, status page or SLA found" ], "where": "local", "x402": "no" }, { "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.", "grade": "D", "name": "Nebius Token Factory fine-tuning", "position": 9, "price": "Pay per use", "score": 47.7, "slug": "nebius-token-factory-fine-tuning", "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" ], "url": "https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning", "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.", "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" ], "where": "hosted", "x402": "no" } ], "updated": "2026-10-08" }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/best/fine-tuning/", "json": "https://www.anchorterminal.com/best/fine-tuning/index.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/best/fine-tuning/index.md", "slim": "https://www.anchorterminal.com/best/fine-tuning/index.min.md" }, "markdown": "All 9 ranked fine-tuning services for AI models on the Anchor benchmark, with a pick for each need and where each one falls short. Scores come from public evidence, re-checked as vendors change.\n\n- Ranked: 9 · agent-ready (BB or better): 1 · accept x402: 0 · hosted endpoints: 6\n- Full ranked table: https://www.anchorterminal.com/categories/fine-tuning.md\n- Head-to-head comparisons: https://www.anchorterminal.com/compare/fine-tuning/index.md (36)\n- Methodology: https://www.anchorterminal.com/benchmark/index.md\n\n## The shortlist\n\n| # | Tool | Grade | Score | Best for | Price | Where |\n| --- | --- | --- | --- | --- | --- | --- |\n| 1 | [Amazon Bedrock model customisation](https://www.anchorterminal.com/tools/amazon-bedrock-customization.md) | BB | 75.8 | Teams already on AWS that want to tune Amazon Nova or Llama models, or run reinforcement fine-tuning with a Lambda reward function, and serve the result inside Bedrock. | Pay per use | hosted |\n| 2 | [Axolotl](https://www.anchorterminal.com/tools/axolotl.md) | B | 64.8 | A team that wants a repeatable, config-driven fine-tune of an open model on its own or rented GPUs, including multi-GPU and multi-node runs. | Free · OSS | library |\n| 3 | [Vertex AI Gemini tuning](https://www.anchorterminal.com/tools/vertex-ai-tuning.md) | B | 64.2 | Teams already on Google Cloud who need to tune Gemini itself, especially with RL, and will serve it there. | Pay per use | hosted |\n| 4 | [Microsoft Foundry fine-tuning (Azure OpenAI)](https://www.anchorterminal.com/tools/azure-foundry-fine-tuning.md) | C | 61.1 | Teams that must tune an OpenAI model, need Azure's compliance and regional controls, and will serve the result on Azure. | Pay per use | hosted |\n| 5 | [Fireworks AI Fine-tuning](https://www.anchorterminal.com/tools/fireworks-fine-tuning.md) | C | 59 | Teams that want managed SFT, DPO or RFT on large open models and may later write a custom RL loop on the same platform. | Pay per use | hosted |\n| 6 | [Together AI Fine-tuning](https://www.anchorterminal.com/tools/together-fine-tuning.md) | C | 54.7 | Teams that want to tune a large open model, possibly with full fine-tuning, and take the weights away. | Pay per use | hosted |\n| 7 | [Unsloth](https://www.anchorterminal.com/tools/unsloth.md) | D | 51.5 | One person or a small team tuning an open model on their own GPU and keeping the weights. | Free · OSS | library |\n| 8 | [Tinker](https://www.anchorterminal.com/tools/tinker.md) | D | 51 | Researchers and teams writing custom post-training loops, especially RL, who want per-token billing and the weights at the end. | Pay per use | local |\n| 9 | [Nebius Token Factory fine-tuning](https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.md) | D | 47.7 | 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. | Pay per use | hosted |\n\n## Picks by need\n\n- Highest score overall: [Amazon Bedrock model customisation](https://www.anchorterminal.com/tools/amazon-bedrock-customization.md), BB, 75.8/100 on the benchmark. Also [Axolotl](https://www.anchorterminal.com/tools/axolotl.md), B, 64.8/100.\n- Maintenance \u0026 community: [Axolotl](https://www.anchorterminal.com/tools/axolotl.md), 88/100 on maintenance \u0026 community, against 45 for the overall leader.\n- Transparency \u0026 trust: [Vertex AI Gemini tuning](https://www.anchorterminal.com/tools/vertex-ai-tuning.md), 88/100 on transparency \u0026 trust, against 83 for the overall leader.\n- Self-hosting under an open licence: [Axolotl](https://www.anchorterminal.com/tools/axolotl.md), self-hosted, Apache-2 licence. Also [Unsloth](https://www.anchorterminal.com/tools/unsloth.md), self-hosted, Apache-2 licence.\n\n## How to choose\n\n- Weight export and licence: Check whether the tuned weights can be downloaded and under what licence, because an agent that must run offline or on your own hardware needs them.\n- Tunable base models and methods: Check which base models can be tuned and whether your chosen method is supported for each model, since method support can differ between them.\n- Training price by method: Check the training price per token for supervised and reinforcement runs separately, because the reinforcement price can differ from the supervised price on the same data.\n- Serving and idle costs: Check the serving price, idle deletion and checkpoint storage fees, because an agent that calls a tuned model only occasionally can end up paying for idle time.\n\n- How the benchmark tests this category: The same small dataset used to tune a comparable open model on each service, then served. We check the job flow, how long training takes, whether the weights can leave, and the training and serving cost.\n\n## Each one in detail\n\n### 1. Amazon Bedrock model customisation, BB 75.8/100\n\nManaged supervised fine-tuning, reinforcement fine-tuning and distillation of Amazon Nova, Meta Llama and selected open-weight models on Amazon Bedrock, run as asynchronous jobs through the Bedrock control-plane API, the AWS SDKs and CLI, or OpenAI-compatible endpoints.\n\n- Verdict: Job creation takes an idempotency token, job lists filter and paginate, and the Service Terms give the customer exclusive use of a tuned model. Jobs run in two US Regions only, weights can't be exported, and the newest customisation API change found dates from 28 May 2026.\n- Choose it for: Teams already on AWS that want to tune Amazon Nova or Llama models, or run reinforcement fine-tuning with a Lambda reward function, and serve the result inside Bedrock.\n- Strength: `CreateModelCustomizationJob` accepts a `clientRequestToken`, so a repeated create after a timeout doesn't start a second job\n- Strength: `ListModelCustomizationJobs` filters by status, name and creation time, sorts, and pages with `maxResults` up to 1,000 and `nextToken`\n- Strength: Service Terms section 50.12.4 gives the customer exclusive use of a customised model and bars third-party model providers from accessing it\n- Weakness: Fine-tuning runs in us-east-1 and us-west-2 only, and each base model in one of them (two for the Titan models)\n- Weakness: No weight export was found in the reviewed documentation, and Service Terms section 50.11 forbids extracting model weights\n- Weakness: Setup needs an IAM service role, `iam:PassRole` and S3 buckets for input and output before the first job\n- Price: Pay per use · Auth: OAuth or key · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/amazon-bedrock-customization.md\n\n### 2. Axolotl, B 64.8/100\n\nOpen-source command-line tool and Python package for fine-tuning open language models from one YAML config, covering LoRA, QLoRA, full fine-tuning, preference tuning and GRPO on the owner's GPUs.\n\n- Verdict: Axolotl runs a whole fine-tuning job from one YAML file and ships a JSON Schema of its config plus bundled agent docs. It is 0.x software with telemetry on by default, no terms or privacy policy, and the owner supplies the GPU.\n- Choose it for: A team that wants a repeatable, config-driven fine-tune of an open model on its own or rented GPUs, including multi-GPU and multi-node runs.\n- Strength: Apache-2.0, free, and the weights stay on the owner's hardware\n- Strength: `axolotl config-schema` prints the full config as JSON Schema, and `axolotl agent-docs` prints bundled Markdown references by topic\n- Strength: SFT, LoRA, QLoRA, DPO, IPO, KTO, ORPO, GRPO and reward modelling from one config format\n- Weakness: Telemetry to PostHog is on by default and delays training start by 10 seconds until the variable is set either way\n- Weakness: No terms of service, privacy policy, legal entity or security.txt found on axolotl.ai\n- Weakness: Version 0.20.0, with removals in minor releases (FSDP1 in 0.20.0, `relora_steps` renamed in 0.17.0 with no shim)\n- Price: Free · OSS · Auth: None · x402: no · Where: library\n- Full assessment: https://www.anchorterminal.com/tools/axolotl.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-axolotl.md\n\n### 3. Vertex AI Gemini tuning, B 64.2/100\n\nSupervised, preference and reinforcement tuning of Gemini, plus supervised tuning of Gemma, Llama and Qwen, on Google Cloud's Gemini Enterprise Agent Platform (the platform formerly called Vertex AI).\n\n- Verdict: Supervised, preference and reinforcement tuning of Gemini, plus supervised tuning of Gemma, Llama and Qwen. No weight export. The tuned model exists only as a Google Cloud endpoint.\n- Choose it for: Teams already on Google Cloud who need to tune Gemini itself, especially with RL, and will serve it there.\n- Strength: Supervised, preference and reinforcement tuning of Gemini, plus supervised tuning of Gemma, Llama and Qwen\n- Strength: No Vertex or Gemini incidents on the Google Cloud status dashboard from July to September 2026\n- Strength: Public proto for GenAiTuningService with filter and pagination on job lists, and docs pages served as Markdown at `.md.txt`\n- Weakness: No weight export. The tuned model exists only as a Google Cloud endpoint\n- Weakness: Tuned Gemini 3 inference costs 1.5x the base model for as long as you serve it\n- Weakness: Setup needs a project, billing, IAM and a Cloud Storage bucket before the first job\n- Price: Pay per use · Auth: OAuth · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/vertex-ai-tuning.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-vertex-ai-tuning.md\n\n### 4. Microsoft Foundry fine-tuning (Azure OpenAI), C 61.1/100\n\nAzure's managed service for supervised, preference and reinforcement fine-tuning of supported OpenAI and open-weight models.\n\n- Verdict: SFT, DPO and RFT on GPT-4.1 and o4-mini through the OpenAI-shaped /openai/v1 API. No weight export; checkpoints copy only between Azure resources.\n- Choose it for: Teams that must tune an OpenAI model, need Azure's compliance and regional controls, and will serve the result on Azure.\n- Strength: SFT, DPO and RFT on GPT-4.1 and o4-mini through the OpenAI-shaped /openai/v1 API\n- Strength: Retirement policy with 60 days' notice and published training and deployment retirement dates per tunable model\n- Strength: Entra ID with RBAC, Azure Monitor logs and an activity log for every customer\n- Weakness: No weight export; checkpoints copy only between Azure resources\n- Weakness: $1.70 an hour hosting on Standard deployments, and deletion after 15 idle days\n- Weakness: GPT-4.1 training at $25 per 1M tokens globally, and no free tier without a card\n- Price: Pay per use · Auth: OAuth or key · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/azure-foundry-fine-tuning.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-azure-foundry-fine-tuning.md\n\n### 5. Fireworks AI Fine-tuning, C 59/100\n\nManaged supervised, preference and reinforcement fine-tuning for open models, with a training API for custom workflows.\n\n- Verdict: SFT, DPO, ORPO and RFT as managed jobs, plus a serverless Training API that is generally available. Tuned LoRAs only deploy to on-demand GPUs at $8 an hour and up, never to serverless.\n- Choose it for: Teams that want managed SFT, DPO or RFT on large open models and may later write a custom RL loop on the same platform.\n- Strength: SFT, DPO, ORPO and RFT as managed jobs, plus a serverless Training API that is generally available\n- Strength: LoRA SFT from $0.50 per 1M training tokens up to 16B parameters, with serving at base-model prices\n- Strength: List endpoints take readMask, pageSize up to 200, AIP-160 filters and orderBy\n- Weakness: Tuned LoRAs only deploy to on-demand GPUs at $8 an hour and up, never to serverless\n- Weakness: No training without a payment method; the $1 sign-up credit buys inference only\n- Weakness: The status page has no training component; a 46-hour cloud provider incident in August hit dedicated deployments\n- Price: Pay per use · Auth: API key · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/fireworks-fine-tuning.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-fireworks-fine-tuning.md\n\n### 6. Together AI Fine-tuning, C 54.7/100\n\nManaged LoRA and full fine-tuning, supervised or DPO, on about 30 open models from Qwen3.5 0.8B to Kimi K2.7, billed per training token with a $4 minimum.\n\n- Verdict: 31 tunable base models, 11 or 12 of them with full fine-tuning as well as LoRA. Fine-tuned models don't run serverless; dedicated endpoints start at $5.49 an hour.\n- Choose it for: Teams that want to tune a large open model, possibly with full fine-tuning, and take the weights away.\n- Strength: 31 tunable base models, 11 or 12 of them with full fine-tuning as well as LoRA\n- Strength: GET /v1/finetune/download returns merged weights or the adapter, at any saved checkpoint\n- Strength: POST /v1/fine-tunes/estimate-price quotes a job before it runs\n- Weakness: Fine-tuned models don't run serverless; dedicated endpoints start at $5.49 an hour\n- Weakness: No free trial, a $5 prepaid purchase before the first call, and job minimums up to $60\n- Weakness: The status page covers serverless models only, and no fine-tuning rate limits are published\n- Price: Pay per use · Auth: API key · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/together-fine-tuning.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-together-fine-tuning.md\n\n### 7. Unsloth, D 51.5/100\n\nOpen-source library, web UI (Studio) and desktop app for LoRA, QLoRA, full fine-tuning and RL (GRPO, DPO, ORPO) of open models on your own GPU, from 3 GB of VRAM.\n\n- Verdict: The Apache-2.0 core runs on customer hardware and keeps model weights there. Users supply and pay for the GPU.\n- Choose it for: One person or a small team tuning an open model on their own GPU and keeping the weights.\n- Strength: Free and open source, Apache-2.0 core, with the weights staying on your hardware\n- Strength: LoRA, QLoRA, full fine-tuning, GRPO, DPO and ORPO from one package\n- Strength: Exports adapters, merged 16-bit weights and GGUF for vLLM, Ollama or llama.cpp\n- Weakness: Not a hosted service; you bring and pay for the GPU\n- Weakness: Studio is AGPL-3.0, and its server-side tools are on by default when exposed\n- Weakness: 792 open issues and 472 open pull requests\n- Price: Free · OSS · Auth: None · x402: no · Where: library\n- Full assessment: https://www.anchorterminal.com/tools/unsloth.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-unsloth.md\n\n### 8. Tinker, D 51/100\n\nThinking Machines Lab's API for model training.\n\n- Verdict: Full control of the training loop with the GPUs abstracted away, plus recipes for SFT, DPO, RL and distillation. LoRA only; no full-parameter training.\n- Choose it for: Researchers and teams writing custom post-training loops, especially RL, who want per-token billing and the weights at the end.\n- Strength: Full control of the training loop with the GPUs abstracted away, plus recipes for SFT, DPO, RL and distillation\n- Strength: Checkpoints download and merge into Hugging Face safetensors, so the weights can leave\n- Strength: Per-token billing with machine-readable prices in models.json\n- Weakness: LoRA only; no full-parameter training\n- Weakness: Python SDK only, with no REST reference or OpenAPI\n- Weakness: No terms of service, status page or SLA found\n- Price: Pay per use · Auth: API key · x402: no · Where: local\n- Full assessment: https://www.anchorterminal.com/tools/tinker.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-tinker.md\n\n### 9. Nebius Token Factory fine-tuning, D 47.7/100\n\nNebius 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.\n\n- 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.\n- Choose it for: 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.\n- Strength: 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- Strength: Checkpoint files download through `GET /v1/files/{file_id}/content`, and an `hf` integration pushes the result to a Hugging Face repository\n- Strength: A public OpenAPI 3.1 file covers the fine-tuning, files, datasets and operations paths, with ranges on every hyperparameter\n- Weakness: 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- Weakness: The models page says deployment is by dedicated endpoints only, and custom model weights are in beta and available on request\n- Weakness: A bank card is mandatory at onboarding, so the $1 trial credit (30 days) is not a card-free trial\n- Price: Pay per use · Auth: API key · x402: no · Where: hosted\n- Full assessment: https://www.anchorterminal.com/tools/nebius-token-factory-fine-tuning.md\n- Against #1: https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-nebius-token-factory-fine-tuning.md\n\n## Head to head\n\n- [Amazon Bedrock model customisation vs Axolotl](https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-axolotl.md)\n- [Amazon Bedrock model customisation vs Vertex AI Gemini tuning](https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-vertex-ai-tuning.md)\n- [Amazon Bedrock model customisation vs Microsoft Foundry fine-tuning (Azure OpenAI)](https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-azure-foundry-fine-tuning.md)\n- [Amazon Bedrock model customisation vs Fireworks AI Fine-tuning](https://www.anchorterminal.com/compare/amazon-bedrock-customization-vs-fireworks-fine-tuning.md)\n- [Axolotl vs Vertex AI Gemini tuning](https://www.anchorterminal.com/compare/axolotl-vs-vertex-ai-tuning.md)\n- [Axolotl vs Microsoft Foundry fine-tuning (Azure OpenAI)](https://www.anchorterminal.com/compare/axolotl-vs-azure-foundry-fine-tuning.md)\n- [Axolotl vs Fireworks AI Fine-tuning](https://www.anchorterminal.com/compare/axolotl-vs-fireworks-fine-tuning.md)\n- [Microsoft Foundry fine-tuning (Azure OpenAI) vs Vertex AI Gemini tuning](https://www.anchorterminal.com/compare/azure-foundry-fine-tuning-vs-vertex-ai-tuning.md)\n- [Fireworks AI Fine-tuning vs Vertex AI Gemini tuning](https://www.anchorterminal.com/compare/fireworks-fine-tuning-vs-vertex-ai-tuning.md)\n- [Microsoft Foundry fine-tuning (Azure OpenAI) vs Fireworks AI Fine-tuning](https://www.anchorterminal.com/compare/azure-foundry-fine-tuning-vs-fireworks-fine-tuning.md)\n\n## Questions\n\n### What are the highest-rated fine-tuning services for AI models for AI agents?\n\nAmazon Bedrock model customisation has the highest benchmark score of the 9 ranked fine-tuning services for AI models, 75.8 (BB). Axolotl is second with 64.8 (B).\n\n### How many fine-tuning services for AI models are agent-ready?\n\n1 of the 9 ranked here grade BB or better, the bar for agent-ready on the Anchor benchmark.\n\n### Which fine-tuning services for AI models accept x402 payments?\n\nNone of the ranked listings here accepts x402 for its main call yet.\n\n### How is this list ranked?\n\nBy the Anchor benchmark score out of 100, a weighted mean of the scored categories minus deductions for negative events, from public evidence re-checked as vendors change. Listings cannot pay for a place. The latest assessment behind this page is from 8 October 2026.\n\n## How this list is made\n\nThe order is the Anchor benchmark score, the same number as on each listing. Each listing is graded from public evidence against the benchmark checklist, and the picks are worked out from those grades, prices and facts. No listing pays for its place, and paid audits or listing help never change a score.\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-10-10", "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": "Best of", "url": "https://www.anchorterminal.com/best/" }, { "name": "Fine-tuning", "url": "" } ], "description": "Amazon Bedrock model customisation (BB), Axolotl (B) and Vertex AI Gemini tuning (B) lead the 9 ranked fine-tuning services for AI models. Picks by need, strengths, weaknesses and prices from the Anchor benchmark.", "facts": [ "Amazon Bedrock model customisation BB", "Axolotl B", "Vertex AI Gemini tuning B" ], "h1": "Best fine-tuning services for AI models", "image": "https://www.anchorterminal.com/assets/og/best-fine-tuning.png", "path": "/best/fine-tuning/", "published": "", "section": "tools", "title": "Best fine-tuning services for AI models in 2026, ranked", "toc": null, "updated": "2026-10-08", "url": "https://www.anchorterminal.com/best/fine-tuning/" }, "tokens": { "markdown": 5300, "slim": 1330 }, "version": 1 }