{ "data": { "similar": [ { "grade": "BB", "json": "https://www.anchorterminal.com/tools/google-cloud-document-ai.json", "name": "Google Cloud Document AI", "score": 73.9, "shared": [ "docs.extract" ], "slug": "google-cloud-document-ai" }, { "grade": "BB", "json": "https://www.anchorterminal.com/tools/amazon-textract.json", "name": "Amazon Textract", "score": 73.5, "shared": [ "docs.extract" ], "slug": "amazon-textract" }, { "grade": "BB", "json": "https://www.anchorterminal.com/tools/openai-decisions-api.json", "name": "OpenAI Decisions API", "score": 71.5, "shared": [ "inference.decision" ], "slug": "openai-decisions-api" }, { "grade": "B", "json": "https://www.anchorterminal.com/tools/decider.json", "name": "Decider", "score": 69.5, "shared": [ "inference.decision" ], "slug": "decider" }, { "grade": "B", "json": "https://www.anchorterminal.com/tools/nace-drex.json", "name": "Drex 1.5", "score": 69.5, "shared": [ "inference.decision" ], "slug": "nace-drex" }, { "grade": "B", "json": "https://www.anchorterminal.com/tools/convai-laya.json", "name": "Laya", "score": 69.2, "shared": [ "inference.decision" ], "slug": "convai-laya" } ], "tool": { "slug": "scaledown", "name": "ScaleDown", "vendor": "Carbon ScaleDown, Inc.", "vendorUrl": "https://scaledown.ai", "kind": "model", "category": "decision-models", "summary": "ScaleDown's API serves three single-task Constrained Decoding Models (Classify, Extract and Summarize) and a Compress model, called with an API key at api.scaledown.xyz. Classify answers Jev-shaped decision requests with probabilities.", "url": "https://www.anchorterminal.com/tools/scaledown", "markdownUrl": "https://www.anchorterminal.com/tools/scaledown.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/scaledown.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/scaledown.json", "repo": "https://github.com/scaledown-team/langchain-scaledown", "license": "Proprietary hosted models under ScaleDown's Terms of Service. The langchain-scaledown integration is MIT", "transports": [ "http" ], "remoteUrl": "https://api.scaledown.xyz", "packages": [ { "registry": "pypi", "name": "langchain-scaledown" } ], "auth": "api-key", "authNotes": "An API key in the `x-api-key` header. Extract and Summarize on `/v1/chat/completions` also accept `Authorization: Bearer`, so an unmodified OpenAI SDK works by changing the base URL. Keys are made in the dashboard after sign-up at scaledown.ai/auth. We found no documentation of key scopes, expiry or rotation. An unauthenticated request returns 401 with `{\"detail\":\"x-api-key header is required\"}`.", "pricing": "usage", "pricingNotes": "$0.05 per million input tokens for every model, with output tokens free (https://scaledown.ai, FAQ). Classify's optional one-sentence reasoning bills its extra input tokens at 2x. New accounts get 50 million free tokens with no card, per the homepage. The docs introduction instead says to contact sales for a key. OpenRouter lists Extract and Summarize at $0.04 per million input tokens. Enterprise and self-hosted plans are priced by sales.", "priceSummary": "Pay per use", "where": "hosted", "x402": { "level": "no", "evidence": "No x402, MPP or L402 in the docs, the OpenAPI document, the site or the LiteLLM and OpenRouter entries (checked 2026-10-14). Running out of credit returns 402 Payment Required as an ordinary error.", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": 0, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-10-14" }, "docsUrl": "https://docs.scaledown.ai/introduction", "llmsTxt": "https://docs.scaledown.ai/llms.txt", "openapi": "https://docs.scaledown.ai/api-reference/openapi.json", "capabilities": [ "inference.decision", "docs.extract" ], "tags": [ "hosted", "model", "closed-source", "openapi", "llms-txt", "openai-compatible", "usage-priced", "free-tier", "no-card", "python" ], "lastRelease": "2026-10-14", "graded": true, "anchor": { "graded": true, "score": 48.6, "grade": "D", "agentReady": false, "rank": 873, "ranked": true, "rankOf": 1084, "categoryRank": 14, "methodology": "0.4", "run": "2026-10-01", "scores": { "ergonomics": 76, "maintenance": 56, "payments": 40, "reliability": 20, "schema": 69, "security": 48, "transparency": 54 }, "pending": [ "performance", "tasks" ], "breakdown": [ { "key": "reliability", "name": "Reliability", "weight": 16, "effectiveWeight": 20, "score": 20, "points": 4, "reason": "Read as a hosted model API. No status page found. status.scaledown.ai and status.scaledown.xyz don't resolve, and the site and docs link to none (0). No readable incident history (5). No rate limit with numbers. The site promises generous fair-use limits and the terms (2.3) let ScaleDown change limits on reasonable notice (0). Only the Compress reference lists 429 with advice to back off and retry. Classify, Decisions, Extract and Summarize list 502 or 500 and 504 without retry guidance. A call has no side effects, so a retry is safe (5 of 15). No SLA published. llms.txt mentions SLAs only for enterprise and self-hosted contracts (0). The models carry no beta or preview label and are listed on OpenRouter and Vercel AI Gateway (10)." }, { "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": 69, "points": 11.21, "reason": "An OpenAPI 3.1 document (ScaleDown API 1.0.0) covers /compress/raw/, /classify and /v1/scaledown. Its /classify schema has only `text` and `labels`, without `chunks`, `document`, `multi_label` or the label caps, and Extract and Summarize on /v1/chat/completions aren't in it (15 of 25). llms.txt indexes 28 pages, each with a Markdown twin (10). Overviews say when to use each model and how to write rubrics, and a table says when to choose /classify over Decisions. No known-limitations page (14 of 20). Typed question kinds behind a `type` field, with 20 labels, 16 in multi-label mode, 50 chunks and 2 to 10 score levels stated in prose, and Extract driven by a JSON schema with `$ref` refused (11 of 15). curl, Python and TypeScript examples with sample responses, and an error table per endpoint (422, 402, 502, 504 and others), though 429 appears only for Compress and the error body is shown only by our 401 (12 of 15). Pinned model IDs (`classify-1`, `extract-1`, `summarize-1`). No public changelog, but the docs source on GitHub has dated commits (7 of 15)." }, { "key": "ergonomics", "name": "Agent ergonomics", "weight": 13, "effectiveWeight": 16.25, "score": 76, "points": 12.35, "reason": "Read as an API an agent calls for a decision, the reading used for Jev. Decisions answers are a few numbers per question, and the reference's sample answers a three-way choice from 62 input tokens with 1 output token. Extract adds 500 characters of context on each side of every value by default, which `context_chars` can turn off, and inputs reach 1M tokens per the site FAQ (20 of 25). `threshold`, `top_n`, `context_chars`, multi-label mode, the `reasoning` switch and up to 50 chunks a request sized against one billed label set (17 of 20). Error tables per endpoint, with 402 for insufficient credit and a plain `detail` message on 401. Error bodies aren't documented (13 of 20). Calls have no side effects. No retry guidance outside Compress, and the langchain-scaledown client doesn't retry (14 of 20). No SDK of ScaleDown's own. OpenAI SDKs work against /v1/chat/completions with only the base URL changed, a Jev client works against /v1/scaledown, and langchain-scaledown, LiteLLM and OpenRouter cover Python and TypeScript callers. `model` is optional on Decisions (12 of 15)." }, { "key": "security", "name": "Security \u0026 auth", "weight": 14, "effectiveWeight": 17.5, "score": 48, "points": 8.4, "reason": "Read for a model API, on the lines used for Jev. A plain API key in `x-api-key`, or as a bearer token on chat completions. We found no scopes, expiry or rotation documented and didn't check revocation, since the dashboard is behind sign-in and disallowed in robots.txt (18 of 30). The terms (6.3), the DPA (20 July 2026) and the FAQ agree that customer data isn't used for training. The DPA says request data is held only in memory during inference, operational logs without customer data are kept up to 90 days, and processing stays in the United States. llms.txt says prompts are never logged, which the DPA's operational logs qualify (17 of 20). Classify and Extract judge text that often comes from users. We found no warning or mitigation for injected instructions (3 of 15). Each Decisions response reports tokens and cost. llms.txt puts audit logs on enterprise plans only (6 of 15). No security.txt (404), disclosure policy or bug bounty found. The DPA commits to breach notice within 48 hours. Certification claims conflict (DPA SOC 2 Type II and ISO 27001, homepage SOC 2 Type 1 badge, llms.txt in progress), and we found no trust centre or report to settle them (4 of 20)." }, { "key": "payments", "name": "Payments \u0026 pricing", "weight": 10, "effectiveWeight": 12.5, "score": 40, "points": 5, "reason": "No x402, MPP or L402 on ScaleDown's endpoints or through the gateways that list it (0). Per-token price published without a login, $0.05 per million input tokens with output free (20). The homepage says new accounts get 50 million free tokens with no credit card. The docs introduction says to contact sales for a key, and we didn't sign up, so the free tier rests on the homepage (20). A person signs up in a browser and creates a key in the dashboard (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": 56, "points": 4.9, "reason": "Read for a closed model, on the lines used for Jev. Classify, Extract and Summarize reached Vercel AI Gateway and OpenRouter on 13 October 2026, and the docs changed on 14 October (30). Pinned `-1` model IDs exist, but the terms (2.2) promise only commercially reasonable efforts to give notice of material adverse changes, and there's no deprecation policy (5 of 12). One version of each model so far. The docs dropped the native /extract and /summarization/abstractive endpoints and the batch pages from the reference on 13 October, while the site and LiteLLM's guide still name those paths (2 of 8). No public changelog. Dated commits in the docs repository, a contact address and a Discord bot repository, and issue replies weren't sampled (6 of 15). No SDK of ScaleDown's own. langchain-scaledown 0.2.0 (MIT, 12 October 2026) is published by the scaledown-team organisation, and OpenAI SDK compatibility covers other languages (8 of 15). langchain-scaledown ships GitHub workflows and unit and integration tests, released two days ago (5 of 10)." }, { "key": "transparency", "name": "Transparency \u0026 trust", "weight": 7, "effectiveWeight": 8.75, "score": 54, "points": 4.72, "note": "editorial 56, provenance 52", "reason": "Closed hosted models under published terms of service (20 July 2026), a privacy notice (3 June 2026) and a DPA, naming Carbon ScaleDown, Inc. and Delaware law. The legal pages render only in the site's JavaScript and the server answers them with 404 (15). The terms, DPA and FAQ agree on no training, and the DPA states retention (none for request data, 90 days for operational logs). llms.txt's claim that nothing is logged and the conflicting certification claims count against agreement (20 of 30). No deprecation policy or dated notices. The terms promise commercially reasonable efforts (5 of 20). The DPA lists four subprocessors (AWS, Google Cloud, Modal Labs, Supabase), all in the United States, and commits to 30 days' notice of any change that would break its retention or training commitments (16 of 20)." } ], "assessment": { "date": "2026-10-14", "basis": "public evidence", "confidence": "medium", "notes": { "ergonomics": "Read as an API an agent calls for a decision, the reading used for Jev. Decisions answers are a few numbers per question, and the reference's sample answers a three-way choice from 62 input tokens with 1 output token. Extract adds 500 characters of context on each side of every value by default, which `context_chars` can turn off, and inputs reach 1M tokens per the site FAQ (20 of 25). `threshold`, `top_n`, `context_chars`, multi-label mode, the `reasoning` switch and up to 50 chunks a request sized against one billed label set (17 of 20). Error tables per endpoint, with 402 for insufficient credit and a plain `detail` message on 401. Error bodies aren't documented (13 of 20). Calls have no side effects. No retry guidance outside Compress, and the langchain-scaledown client doesn't retry (14 of 20). No SDK of ScaleDown's own. OpenAI SDKs work against /v1/chat/completions with only the base URL changed, a Jev client works against /v1/scaledown, and langchain-scaledown, LiteLLM and OpenRouter cover Python and TypeScript callers. `model` is optional on Decisions (12 of 15).", "maintenance": "Read for a closed model, on the lines used for Jev. Classify, Extract and Summarize reached Vercel AI Gateway and OpenRouter on 13 October 2026, and the docs changed on 14 October (30). Pinned `-1` model IDs exist, but the terms (2.2) promise only commercially reasonable efforts to give notice of material adverse changes, and there's no deprecation policy (5 of 12). One version of each model so far. The docs dropped the native /extract and /summarization/abstractive endpoints and the batch pages from the reference on 13 October, while the site and LiteLLM's guide still name those paths (2 of 8). No public changelog. Dated commits in the docs repository, a contact address and a Discord bot repository, and issue replies weren't sampled (6 of 15). No SDK of ScaleDown's own. langchain-scaledown 0.2.0 (MIT, 12 October 2026) is published by the scaledown-team organisation, and OpenAI SDK compatibility covers other languages (8 of 15). langchain-scaledown ships GitHub workflows and unit and integration tests, released two days ago (5 of 10).", "payments": "No x402, MPP or L402 on ScaleDown's endpoints or through the gateways that list it (0). Per-token price published without a login, $0.05 per million input tokens with output free (20). The homepage says new accounts get 50 million free tokens with no credit card. The docs introduction says to contact sales for a key, and we didn't sign up, so the free tier rests on the homepage (20). A person signs up in a browser and creates a key in the dashboard (0).", "reliability": "Read as a hosted model API. No status page found. status.scaledown.ai and status.scaledown.xyz don't resolve, and the site and docs link to none (0). No readable incident history (5). No rate limit with numbers. The site promises generous fair-use limits and the terms (2.3) let ScaleDown change limits on reasonable notice (0). Only the Compress reference lists 429 with advice to back off and retry. Classify, Decisions, Extract and Summarize list 502 or 500 and 504 without retry guidance. A call has no side effects, so a retry is safe (5 of 15). No SLA published. llms.txt mentions SLAs only for enterprise and self-hosted contracts (0). The models carry no beta or preview label and are listed on OpenRouter and Vercel AI Gateway (10).", "schema": "An OpenAPI 3.1 document (ScaleDown API 1.0.0) covers /compress/raw/, /classify and /v1/scaledown. Its /classify schema has only `text` and `labels`, without `chunks`, `document`, `multi_label` or the label caps, and Extract and Summarize on /v1/chat/completions aren't in it (15 of 25). llms.txt indexes 28 pages, each with a Markdown twin (10). Overviews say when to use each model and how to write rubrics, and a table says when to choose /classify over Decisions. No known-limitations page (14 of 20). Typed question kinds behind a `type` field, with 20 labels, 16 in multi-label mode, 50 chunks and 2 to 10 score levels stated in prose, and Extract driven by a JSON schema with `$ref` refused (11 of 15). curl, Python and TypeScript examples with sample responses, and an error table per endpoint (422, 402, 502, 504 and others), though 429 appears only for Compress and the error body is shown only by our 401 (12 of 15). Pinned model IDs (`classify-1`, `extract-1`, `summarize-1`). No public changelog, but the docs source on GitHub has dated commits (7 of 15).", "security": "Read for a model API, on the lines used for Jev. A plain API key in `x-api-key`, or as a bearer token on chat completions. We found no scopes, expiry or rotation documented and didn't check revocation, since the dashboard is behind sign-in and disallowed in robots.txt (18 of 30). The terms (6.3), the DPA (20 July 2026) and the FAQ agree that customer data isn't used for training. The DPA says request data is held only in memory during inference, operational logs without customer data are kept up to 90 days, and processing stays in the United States. llms.txt says prompts are never logged, which the DPA's operational logs qualify (17 of 20). Classify and Extract judge text that often comes from users. We found no warning or mitigation for injected instructions (3 of 15). Each Decisions response reports tokens and cost. llms.txt puts audit logs on enterprise plans only (6 of 15). No security.txt (404), disclosure policy or bug bounty found. The DPA commits to breach notice within 48 hours. Certification claims conflict (DPA SOC 2 Type II and ISO 27001, homepage SOC 2 Type 1 badge, llms.txt in progress), and we found no trust centre or report to settle them (4 of 20).", "transparency": "Closed hosted models under published terms of service (20 July 2026), a privacy notice (3 June 2026) and a DPA, naming Carbon ScaleDown, Inc. and Delaware law. The legal pages render only in the site's JavaScript and the server answers them with 404 (15). The terms, DPA and FAQ agree on no training, and the DPA states retention (none for request data, 90 days for operational logs). llms.txt's claim that nothing is logged and the conflicting certification claims count against agreement (20 of 30). No deprecation policy or dated notices. The terms promise commercially reasonable efforts (5 of 20). The DPA lists four subprocessors (AWS, Google Cloud, Modal Labs, Supabase), all in the United States, and commits to 30 days' notice of any change that would break its retention or training commitments (16 of 20)." }, "sources": [ { "what": "docs index (llms.txt)", "url": "https://docs.scaledown.ai/llms.txt", "seen": "2026-10-14" }, { "what": "docs introduction", "url": "https://docs.scaledown.ai/introduction.md", "seen": "2026-10-14" }, { "what": "Classify overview", "url": "https://docs.scaledown.ai/api-reference/classify-overview.md", "seen": "2026-10-14" }, { "what": "Classify API reference", "url": "https://docs.scaledown.ai/api-reference/classify.md", "seen": "2026-10-14" }, { "what": "Decisions overview", "url": "https://docs.scaledown.ai/api-reference/decisions-overview.md", "seen": "2026-10-14" }, { "what": "Decisions API reference", "url": "https://docs.scaledown.ai/api-reference/decisions.md", "seen": "2026-10-14" }, { "what": "Extract overview", "url": "https://docs.scaledown.ai/api-reference/extract-overview.md", "seen": "2026-10-14" }, { "what": "Extract API reference", "url": "https://docs.scaledown.ai/api-reference/extract.md", "seen": "2026-10-14" }, { "what": "Summarize overview", "url": "https://docs.scaledown.ai/api-reference/summarize-overview.md", "seen": "2026-10-14" }, { "what": "Summarize API reference", "url": "https://docs.scaledown.ai/api-reference/summarize.md", "seen": "2026-10-14" }, { "what": "Compress API reference (429 handling)", "url": "https://docs.scaledown.ai/api-reference/compress.md", "seen": "2026-10-14" }, { "what": "OpenAPI document", "url": "https://docs.scaledown.ai/api-reference/openapi.json", "seen": "2026-10-14" }, { "what": "benchmarks overview", "url": "https://docs.scaledown.ai/benchmarks/overview.md", "seen": "2026-10-14" }, { "what": "site llms.txt (pricing, compliance, benchmarks)", "url": "https://scaledown.ai/llms.txt", "seen": "2026-10-14" }, { "what": "homepage, FAQ, terms, privacy notice and DPA as served in the site bundle", "url": "https://scaledown.ai/assets/index-CIrIjse9.js", "seen": "2026-10-14" }, { "what": "terms of service", "url": "https://scaledown.ai/terms", "seen": "2026-10-14" }, { "what": "privacy notice", "url": "https://scaledown.ai/privacy", "seen": "2026-10-14" }, { "what": "data processing addendum", "url": "https://scaledown.ai/dpa", "seen": "2026-10-14" }, { "what": "site robots.txt", "url": "https://scaledown.ai/robots.txt", "seen": "2026-10-14" }, { "what": "docs source and commit history", "url": "https://github.com/scaledown-team/docs", "seen": "2026-10-14" }, { "what": "langchain-scaledown source", "url": "https://github.com/scaledown-team/langchain-scaledown", "seen": "2026-10-14" }, { "what": "PyPI langchain-scaledown", "url": "https://pypi.org/project/langchain-scaledown/", "seen": "2026-10-14" }, { "what": "OpenRouter model list", "url": "https://openrouter.ai/api/v1/models", "seen": "2026-10-14" }, { "what": "Vercel AI Gateway model list", "url": "https://ai-gateway.vercel.sh/v1/models", "seen": "2026-10-14" }, { "what": "LiteLLM provider guide", "url": "https://docs.litellm.ai/docs/providers/scaledown", "seen": "2026-10-14" }, { "what": "LangChain launch blog", "url": "https://www.langchain.com/blog/building-long-running-agents-with-scaledown-and-deep-agents", "seen": "2026-10-14" }, { "what": "unauthenticated API response (401)", "url": "https://api.scaledown.xyz/v1/models", "seen": "2026-10-14" }, { "what": "RDAP for scaledown.ai", "url": "https://rdap.org/domain/scaledown.ai", "seen": "2026-10-14" } ], "openQuestions": [ "unchecked: the sign-up flow, the dashboard and whether keys can be scoped, rotated or revoked, since /dashboard, /settings and /getapikey are disallowed in robots.txt and we made no account", "Whether the free tier needs a card. The homepage says no card and 50 million free tokens. The docs introduction says to contact sales. The lead's 'first 50 hours' limit wasn't found on any ScaleDown page", "Which compliance claim is current (SOC 2 Type II and ISO 27001 in the DPA, SOC 2 Type 1 badge on the homepage, in progress in llms.txt)", "Whether the native /extract and /summarization/abstractive endpoints still answer after the docs dropped them on 13 October 2026. LiteLLM's guide and the site still use them", "Rate limits, whether 429 carries Retry-After, and whether failed calls are charged", "The lead's DecisionBench figures (77.2 per cent against Jev's 72 per cent) weren't found on ScaleDown's site, docs or the LangChain post", "TrustedRouter, RouterPlus and an Arthur AI integration weren't confirmed on a first-hand page. Arthur appears only as a logo on the homepage", "Classify isn't in OpenRouter's model list and Extract and Summarize aren't in Vercel AI Gateway's, so the gateways each carry part of the family" ] }, "negative": -2, "negativeNotes": [ "Conflicting compliance claims on ScaleDown's own pages, read 14 October 2026. The DPA (20 July 2026) says ScaleDown maintains a SOC 2 Type II attestation and ISO 27001 certification, the homepage shows SOC 2 Type 1, HIPAA and ISO 27001 badges, and llms.txt says GDPR, HIPAA and SOC 2 compliance is in progress. At least one is wrong. No report or trust centre found to settle it (-2) (https://scaledown.ai/dpa, https://scaledown.ai/llms.txt)" ], "verdict": "Classify answers Jev-shaped `choice`, `noul` and `score` questions with probabilities, and Extract returns values with confidences and character offsets, at $0.05 per million input tokens with output free. There is no status page, no published rate limit and no public changelog, and the site's compliance claims contradict each other.", "bestFor": "Cheap, high-volume labelling, routing and yes or no checks where a probability is more useful than prose, and for pulling described fields with offsets from text, images or PDFs.", "strengths": [ "`POST /v1/scaledown` takes Jev's request shape, so a Jev integration can move by changing the base URL, per the docs", "$0.05 per million input tokens for every model, output free, and 50 million free tokens with no card, per the homepage", "Extract returns each value with a confidence, character offsets and surrounding context, and routes `enum` fields to the Classify model", "llms.txt with Markdown twins, an OpenAPI 3.1 document and OpenAI-compatible chat completions for Extract and Summarize", "The terms, DPA and FAQ agree that customer data isn't used for training, and the DPA lists four US subprocessors" ], "weaknesses": [ "No status page, no published rate limits and no SLA outside enterprise contracts", "No public changelog. The docs dropped the native /extract and /summarization endpoints from the reference on 13 October 2026", "Compliance claims disagree between the DPA (SOC 2 Type II, ISO 27001), the homepage badges (SOC 2 Type 1) and llms.txt (in progress)", "No security.txt, disclosure policy or bug bounty, and no guidance on hostile text in classified input", "The OpenAPI document covers /classify, /v1/scaledown and Compress only, without the documented limits" ], "agentNotes": [ "Send several questions about one item in one `/v1/scaledown` call. Each becomes its own Classify call and the usage is summed", "Pin `classify-1`, `extract-1` or `summarize-1` once you've tuned thresholds on your own labelled cases", "Set `context_chars` to 0 and use `top_n` on Extract unless you need the 500 characters of context each side", "Set `reasoning: false` on /classify when you only need the label. The reasoning call bills its input at 2x", "Back off and retry on 429, 502 and 504. No limits are published and calls have no side effects" ], "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": 48.6 } ], "editorialScores": { "ergonomics": 76, "maintenance": 56, "payments": 40, "reliability": 20, "schema": 69, "security": 48, "transparency": 56 }, "provenanceScore": 52 }, "connect": { "install": "pip install langchain-scaledown # LangChain middleware; any OpenAI SDK also works against /v1/chat/completions", "http": "curl https://api.scaledown.xyz/v1/scaledown \\\n -H \"x-api-key: $SCALEDOWN_API_KEY\" -H \"Content-Type: application/json\" \\\n -d '{\"model\":\"classify-1\",\"state\":\"I was charged twice for my subscription this month.\",\"questions\":{\"category\":{\"type\":\"choice\",\"criteria\":{\"billing\":\"About a charge, invoice, refund, or payment problem.\",\"technical\":\"About a bug or something not working.\"}}}}'" }, "letme": { "capability": "https://letme.dev/inference.decision", "tool": "https://letme.dev/scaledown" }, "alsoIn": [ "document-extraction" ], "notable": [ "The docs call Classify a Constrained Decoding Model (CDM) for classification and structured decisions, and Extract routes `enum` and `oneOf` keys to it (https://docs.scaledown.ai/introduction.md, https://docs.scaledown.ai/api-reference/extract-overview.md)", "`POST /v1/scaledown` takes TypeSafe Jev's request shape (`state` plus `choice`, `noul` and `score` questions) and is documented as a drop-in replacement, running each question as an internal Classify call (https://docs.scaledown.ai/api-reference/decisions-overview.md)", "The site FAQ says Compress, Extract and Classify are encoder-only and Summarize has a generative decoder step, and that Summarize, Extract and Classify each take a 1M-token context (https://scaledown.ai)", "The homepage claims 467x lower cost and 200x higher speed than frontier models. These are ScaleDown's own figures, and its benchmark pages compare against public datasets (FinanceBench, QMSum, CUAD, CLINC OOS), not DecisionBench (https://scaledown.ai/llms.txt)", "On 13 October 2026 Vercel AI Gateway listed `scaledown/scaledown-classify-1` at $0.05 per million input tokens, and OpenRouter listed `scaledown/extract` and `scaledown/summarize` at $0.04 with a 1,000,000-token context (https://ai-gateway.vercel.sh/v1/models, https://openrouter.ai/api/v1/models)", "LiteLLM documents five ScaleDown routes (classify, extract, summarize, decisions, compress) and rejects images, files and `strict: true` schemas (https://docs.litellm.ai/docs/providers/scaledown)", "LangChain's blog of 13 October 2026, co-written with ScaleDown's CEO, introduces langchain-scaledown middleware and a `/scaledown` path in the LangSmith LLM Gateway (https://www.langchain.com/blog/building-long-running-agents-with-scaledown-and-deep-agents)", "The site's compliance claims disagree. The DPA says ScaleDown maintains SOC 2 Type II and ISO 27001, the homepage shows SOC 2 Type 1, HIPAA and ISO 27001 badges, and llms.txt says GDPR, HIPAA and SOC 2 compliance is in progress (https://scaledown.ai/dpa, https://scaledown.ai/llms.txt)", "The scaledown.ai robots.txt allows general crawlers but sets `Content-Signal: ai-train=no, search=yes, ai-input=no` and disallows GPTBot, Claude-Web and anthropic-ai. The docs site sets ai-input=yes (https://scaledown.ai/robots.txt, https://docs.scaledown.ai/robots.txt)" ], "area": "models", "details": [ { "label": "Models", "value": "classify and classify-1 (Classify, also behind the Jev-shaped Decisions API), extract and extract-1, summarize and summarize-1, and compress. The bare names follow the latest version and the -1 names pin version 1" }, { "label": "Endpoints", "value": "POST /classify and POST /v1/scaledown for Classify, POST /v1/chat/completions for Extract and Summarize, POST /compress/raw/ for Compress, GET /v1/models" }, { "label": "Question types", "value": "choice (one option, probabilities sum to 1), noul (probability of yes), score (2 to 10 ordered levels, returned as an expected value). /classify takes up to 20 labels with rubrics, 16 in multi-label mode, and 50 chunks a request" }, { "label": "Extract output", "value": "Each value with a confidence, start and end character offsets and 500 characters of context each side by default, up to 30 values per type, plus nested objects and arrays" }, { "label": "Summarize output", "value": "At most 1,000 output tokens, with finish_reason length when cut off. Sampling parameters are fixed" }, { "label": "Context and inputs", "value": "1M tokens for Summarize, Extract and Classify per the site FAQ and OpenRouter. Text, or base64 JPEG, PNG, TIFF and PDF documents" }, { "label": "Rate limits", "value": "Not published. The site says generous fair-use limits and the terms let ScaleDown change limits on reasonable notice" }, { "label": "Trains on API data", "value": "No, per the terms (6.3), the DPA and the site FAQ" }, { "label": "Data retention", "value": "Zero retention of request data per the DPA, with operational logs without customer data kept up to 90 days. Processing in the United States only" }, { "label": "Integrations", "value": "OpenRouter (Extract, Summarize), Vercel AI Gateway (Classify-1), LiteLLM, langchain-scaledown 0.2.0 and the LangSmith LLM Gateway" }, { "label": "Status page", "value": "None found" } ], "unitPrices": [ { "item": "Classify, Extract, Summarize and Compress input", "unit": "1m-tokens", "usd": 0.05, "note": "Output tokens free. Classify reasoning input billed at 2x. 50M free tokens for new accounts" } ], "provenance": { "legalEntity": "Carbon ScaleDown, Inc.", "domain": "scaledown.ai", "domainRegistered": "2025-03-28", "endpointOnVendorDomain": true, "terms": "https://scaledown.ai/terms", "privacy": "https://scaledown.ai/privacy", "statusPage": "", "changelog": "", "securityTxt": "none", "checked": "2026-10-14", "notes": [ "The terms (last updated 20 July 2026), privacy notice (3 June 2026) and DPA (20 July 2026) name Carbon ScaleDown, Inc., 22952 Stokes Rd, Ramona, CA 92065, and Delaware law. One sentence of the terms names ScaleDown, Inc.", "scaledown.ai is a single-page app. /terms, /privacy and /dpa answer HTTP 404 with the app shell, and their text is in the JavaScript bundle (https://scaledown.ai/assets/index-CIrIjse9.js), where we read it.", "The API host is api.scaledown.xyz. scaledown.xyz redirects to scaledown.ai, so we treat the endpoint as on the vendor's domain.", "scaledown.ai/.well-known/security.txt and docs.scaledown.ai/.well-known/security.txt return 404. status.scaledown.ai and status.scaledown.xyz don't resolve.", "No public changelog. The docs source at github.com/scaledown-team/docs has dated commits that record API changes.", "RDAP gives scaledown.ai a registration date of 28 March 2025." ], "score": 52, "checks": [ { "check": "Legal entity named", "value": "Carbon ScaleDown, Inc.", "points": 20, "max": 20, "state": "ok" }, { "check": "Domain age", "value": "scaledown.ai, registered 2025-03-28 (1 year)", "points": 3, "max": 15, "state": "part" }, { "check": "Endpoint on the vendor's domain", "value": "api.scaledown.xyz", "points": 15, "max": 15, "state": "ok" }, { "check": "Terms of service", "value": "published, but our reader couldn't read it", "points": 7, "max": 10, "state": "part" }, { "check": "Privacy policy", "value": "published, but our reader couldn't read it", "points": 7, "max": 10, "state": "part" }, { "check": "Status page", "value": "not found", "points": 0, "max": 10, "state": "no" }, { "check": "Changelog", "value": "not found", "points": 0, "max": 10, "state": "no" }, { "check": "security.txt", "value": "not found", "points": 0, "max": 10, "state": "no" } ], "policies": [ { "kind": "terms", "url": "https://scaledown.ai/terms", "state": "unreadable", "reason": "the page answered HTTP 404 to our reader", "readAt": "2026-10-14", "points": 7, "max": 10 }, { "kind": "privacy", "url": "https://scaledown.ai/privacy", "state": "unreadable", "reason": "the page answered HTTP 404 to our reader", "readAt": "2026-10-14", "points": 7, "max": 10 } ] }, "pageJsonUrl": "https://www.anchorterminal.com/tools/scaledown.json", "live": { "slug": "scaledown", "probe": { "target": "https://api.scaledown.xyz", "method": "get", "lastAt": "2026-10-14T22:40:24.725315274Z", "lastOk": true, "lastStatus": 403, "lastMs": 322, "lastNote": "asks for credentials", "authRequired": true, "uptime24h": 100, "uptime30d": 100, "p50ms24h": 267, "p95ms24h": 325, "samples24h": 34, "samples30d": 34, "days": [ { "date": "2026-10-14", "probes": 34, "ok": 34 } ] }, "pages": [ { "url": "https://scaledown.ai/privacy", "kind": "privacy", "status": 404, "checkedAt": "2026-10-14T21:06:40.247126332Z", "changedAt": "0001-01-01T00:00:00Z" }, { "url": "https://scaledown.ai/terms", "kind": "terms", "status": 404, "checkedAt": "2026-10-14T21:06:42.626164273Z", "changedAt": "0001-01-01T00:00:00Z" } ], "updatedAt": "2026-10-14T22:40:24.725315274Z" } }, "verify": { "accepts": "a page on scaledown.ai or one of its subdomains, or the README of github.com/scaledown-team/langchain-scaledown", "badgeUrl": "https://www.anchorterminal.com/badges/scaledown.svg", "body": { "slug": "scaledown", "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/scaledown", "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/scaledown\"\u003e\u003cimg src=\"https://www.anchorterminal.com/badges/scaledown.svg\" alt=\"ScaleDown on Anchor Terminal\" height=\"20\"\u003e\u003c/a\u003e", "markdown": "[![ScaleDown on Anchor Terminal](https://www.anchorterminal.com/badges/scaledown.svg)](https://www.anchorterminal.com/tools/scaledown)", "link": "\u003ca href=\"https://www.anchorterminal.com/tools/scaledown\"\u003eScaleDown 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/scaledown", "json": "https://www.anchorterminal.com/tools/scaledown.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/tools/scaledown.md", "slim": "https://www.anchorterminal.com/tools/scaledown.min.md" }, "markdown": "## Overview\n\n**Grade D · 48.6/100 · rank #873 of 1084 · #14 in Decision models · not agent-ready · confidence medium**\n\n\nAlso listed in [Document parsing \u0026 extraction](https://www.anchorterminal.com/categories/document-extraction.md).\n\n## Assessment\n\nClassify answers Jev-shaped `choice`, `noul` and `score` questions with probabilities, and Extract returns values with confidences and character offsets, at $0.05 per million input tokens with output free. There is no status page, no published rate limit and no public changelog, and the site's compliance claims contradict each other.\n\n## Facts\n\n| Field | Value |\n| --- | --- |\n| Vendor | Carbon ScaleDown, Inc. (https://scaledown.ai) |\n| Kind | Model API |\n| Category | Decision models (https://www.anchorterminal.com/categories/decision-models) |\n| Transport | HTTP |\n| Endpoint | `https://api.scaledown.xyz` |\n| Auth | API key · An API key in the `x-api-key` header. Extract and Summarize on `/v1/chat/completions` also accept `Authorization: Bearer`, so an unmodified OpenAI SDK works by changing the base URL. Keys are made in the dashboard after sign-up at scaledown.ai/auth. We found no documentation of key scopes, expiry or rotation. An unauthenticated request returns 401 with `{\"detail\":\"x-api-key header is required\"}`. |\n| Pricing | Pay per use (Pay per use) · $0.05 per million input tokens for every model, with output tokens free (https://scaledown.ai, FAQ). Classify's optional one-sentence reasoning bills its extra input tokens at 2x. New accounts get 50 million free tokens with no card, per the homepage. The docs introduction instead says to contact sales for a key. OpenRouter lists Extract and Summarize at $0.04 per million input tokens. Enterprise and self-hosted plans are priced by sales. |\n| x402 | No · No x402, MPP or L402 in the docs, the OpenAPI document, the site or the LiteLLM and OpenRouter entries (checked 2026-10-14). Running out of credit returns 402 Payment Required as an ordinary error. |\n| Licence | Proprietary hosted models under ScaleDown's Terms of Service. The langchain-scaledown integration is MIT |\n| Packages | pypi: `langchain-scaledown` |\n| Source | https://github.com/scaledown-team/langchain-scaledown |\n| Docs | https://docs.scaledown.ai/introduction |\n| llms.txt | https://docs.scaledown.ai/llms.txt |\n| Last release | 2026-10-14 |\n| GitHub stars | 0 (as of 2026-10-14) |\n| Models | classify and classify-1 (Classify, also behind the Jev-shaped Decisions API), extract and extract-1, summarize and summarize-1, and compress. The bare names follow the latest version and the -1 names pin version 1 |\n| Endpoints | POST /classify and POST /v1/scaledown for Classify, POST /v1/chat/completions for Extract and Summarize, POST /compress/raw/ for Compress, GET /v1/models |\n| Question types | choice (one option, probabilities sum to 1), noul (probability of yes), score (2 to 10 ordered levels, returned as an expected value). /classify takes up to 20 labels with rubrics, 16 in multi-label mode, and 50 chunks a request |\n| Extract output | Each value with a confidence, start and end character offsets and 500 characters of context each side by default, up to 30 values per type, plus nested objects and arrays |\n| Summarize output | At most 1,000 output tokens, with finish_reason length when cut off. Sampling parameters are fixed |\n| Context and inputs | 1M tokens for Summarize, Extract and Classify per the site FAQ and OpenRouter. Text, or base64 JPEG, PNG, TIFF and PDF documents |\n| Rate limits | Not published. The site says generous fair-use limits and the terms let ScaleDown change limits on reasonable notice |\n| Trains on API data | No, per the terms (6.3), the DPA and the site FAQ |\n| Data retention | Zero retention of request data per the DPA, with operational logs without customer data kept up to 90 days. Processing in the United States only |\n| Integrations | OpenRouter (Extract, Summarize), Vercel AI Gateway (Classify-1), LiteLLM, langchain-scaledown 0.2.0 and the LangSmith LLM Gateway |\n| Status page | None found |\n| Capabilities | inference.decision, docs.extract |\n| Tags | hosted, model, closed-source, openapi, llms-txt, openai-compatible, usage-priced, free-tier, no-card, python |\n| JSON | https://www.anchorterminal.com/api/v1/tools/scaledown.json |\n\n## Score breakdown (methodology v0.4, October 2026 research run)\n\nAssessed 2026-10-14 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 | 20 | 4.0 |\n| Performance | 10% | pending | pending | n/a |\n| Schema \u0026 documentation | 13% | 16.2 | 69 | 11.2 |\n| Agent ergonomics | 13% | 16.2 | 76 | 12.3 |\n| Security \u0026 auth | 14% | 17.5 | 48 | 8.4 |\n| Payments \u0026 pricing | 10% | 12.5 | 40 | 5.0 |\n| Task success | 10% | pending | pending | n/a |\n| Maintenance \u0026 community | 7% | 8.8 | 56 | 4.9 |\n| Transparency \u0026 trust (editorial 56, provenance 52) | 7% | 8.8 | 54 | 4.7 |\n| Negative events | up to −15 | up to −15 | Conflicting compliance claims on ScaleDown's own pages, read 14 October 2026. The DPA (20 July 2026) says ScaleDown maintains a SOC 2 Type II attestation and ISO 27001 certification, the homepage shows SOC 2 Type 1, HIPAA and ISO 27001 badges, and llms.txt says GDPR, HIPAA and SOC 2 compliance is in progress. At least one is wrong. No report or trust centre found to settle it (-2) (https://scaledown.ai/dpa, https://scaledown.ai/llms.txt) | -2 |\n| **Total** | | | | **48.6 → D** |\n\n### Why each score\n\n- Reliability 20: Read as a hosted model API. No status page found. status.scaledown.ai and status.scaledown.xyz don't resolve, and the site and docs link to none (0). No readable incident history (5). No rate limit with numbers. The site promises generous fair-use limits and the terms (2.3) let ScaleDown change limits on reasonable notice (0). Only the Compress reference lists 429 with advice to back off and retry. Classify, Decisions, Extract and Summarize list 502 or 500 and 504 without retry guidance. A call has no side effects, so a retry is safe (5 of 15). No SLA published. llms.txt mentions SLAs only for enterprise and self-hosted contracts (0). The models carry no beta or preview label and are listed on OpenRouter and Vercel AI Gateway (10).\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 69: An OpenAPI 3.1 document (ScaleDown API 1.0.0) covers /compress/raw/, /classify and /v1/scaledown. Its /classify schema has only `text` and `labels`, without `chunks`, `document`, `multi_label` or the label caps, and Extract and Summarize on /v1/chat/completions aren't in it (15 of 25). llms.txt indexes 28 pages, each with a Markdown twin (10). Overviews say when to use each model and how to write rubrics, and a table says when to choose /classify over Decisions. No known-limitations page (14 of 20). Typed question kinds behind a `type` field, with 20 labels, 16 in multi-label mode, 50 chunks and 2 to 10 score levels stated in prose, and Extract driven by a JSON schema with `$ref` refused (11 of 15). curl, Python and TypeScript examples with sample responses, and an error table per endpoint (422, 402, 502, 504 and others), though 429 appears only for Compress and the error body is shown only by our 401 (12 of 15). Pinned model IDs (`classify-1`, `extract-1`, `summarize-1`). No public changelog, but the docs source on GitHub has dated commits (7 of 15).\n- Agent ergonomics 76: Read as an API an agent calls for a decision, the reading used for Jev. Decisions answers are a few numbers per question, and the reference's sample answers a three-way choice from 62 input tokens with 1 output token. Extract adds 500 characters of context on each side of every value by default, which `context_chars` can turn off, and inputs reach 1M tokens per the site FAQ (20 of 25). `threshold`, `top_n`, `context_chars`, multi-label mode, the `reasoning` switch and up to 50 chunks a request sized against one billed label set (17 of 20). Error tables per endpoint, with 402 for insufficient credit and a plain `detail` message on 401. Error bodies aren't documented (13 of 20). Calls have no side effects. No retry guidance outside Compress, and the langchain-scaledown client doesn't retry (14 of 20). No SDK of ScaleDown's own. OpenAI SDKs work against /v1/chat/completions with only the base URL changed, a Jev client works against /v1/scaledown, and langchain-scaledown, LiteLLM and OpenRouter cover Python and TypeScript callers. `model` is optional on Decisions (12 of 15).\n- Security \u0026 auth 48: Read for a model API, on the lines used for Jev. A plain API key in `x-api-key`, or as a bearer token on chat completions. We found no scopes, expiry or rotation documented and didn't check revocation, since the dashboard is behind sign-in and disallowed in robots.txt (18 of 30). The terms (6.3), the DPA (20 July 2026) and the FAQ agree that customer data isn't used for training. The DPA says request data is held only in memory during inference, operational logs without customer data are kept up to 90 days, and processing stays in the United States. llms.txt says prompts are never logged, which the DPA's operational logs qualify (17 of 20). Classify and Extract judge text that often comes from users. We found no warning or mitigation for injected instructions (3 of 15). Each Decisions response reports tokens and cost. llms.txt puts audit logs on enterprise plans only (6 of 15). No security.txt (404), disclosure policy or bug bounty found. The DPA commits to breach notice within 48 hours. Certification claims conflict (DPA SOC 2 Type II and ISO 27001, homepage SOC 2 Type 1 badge, llms.txt in progress), and we found no trust centre or report to settle them (4 of 20).\n- Payments \u0026 pricing 40: No x402, MPP or L402 on ScaleDown's endpoints or through the gateways that list it (0). Per-token price published without a login, $0.05 per million input tokens with output free (20). The homepage says new accounts get 50 million free tokens with no credit card. The docs introduction says to contact sales for a key, and we didn't sign up, so the free tier rests on the homepage (20). A person signs up in a browser and creates a key in the dashboard (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 56: Read for a closed model, on the lines used for Jev. Classify, Extract and Summarize reached Vercel AI Gateway and OpenRouter on 13 October 2026, and the docs changed on 14 October (30). Pinned `-1` model IDs exist, but the terms (2.2) promise only commercially reasonable efforts to give notice of material adverse changes, and there's no deprecation policy (5 of 12). One version of each model so far. The docs dropped the native /extract and /summarization/abstractive endpoints and the batch pages from the reference on 13 October, while the site and LiteLLM's guide still name those paths (2 of 8). No public changelog. Dated commits in the docs repository, a contact address and a Discord bot repository, and issue replies weren't sampled (6 of 15). No SDK of ScaleDown's own. langchain-scaledown 0.2.0 (MIT, 12 October 2026) is published by the scaledown-team organisation, and OpenAI SDK compatibility covers other languages (8 of 15). langchain-scaledown ships GitHub workflows and unit and integration tests, released two days ago (5 of 10).\n- Transparency \u0026 trust 54: Closed hosted models under published terms of service (20 July 2026), a privacy notice (3 June 2026) and a DPA, naming Carbon ScaleDown, Inc. and Delaware law. The legal pages render only in the site's JavaScript and the server answers them with 404 (15). The terms, DPA and FAQ agree on no training, and the DPA states retention (none for request data, 90 days for operational logs). llms.txt's claim that nothing is logged and the conflicting certification claims count against agreement (20 of 30). No deprecation policy or dated notices. The terms promise commercially reasonable efforts (5 of 20). The DPA lists four subprocessors (AWS, Google Cloud, Modal Labs, Supabase), all in the United States, and commits to 30 days' notice of any change that would break its retention or training commitments (16 of 20).\n\nFix list for a coding agent, everything this grade says the listing lacks, the biggest gain first (22 items): https://www.anchorterminal.com/fixes/scaledown.md (JSON https://www.anchorterminal.com/fixes/scaledown.json)\n\n### What we couldn't check\n\n- unchecked: the sign-up flow, the dashboard and whether keys can be scoped, rotated or revoked, since /dashboard, /settings and /getapikey are disallowed in robots.txt and we made no account\n- Whether the free tier needs a card. The homepage says no card and 50 million free tokens. The docs introduction says to contact sales. The lead's 'first 50 hours' limit wasn't found on any ScaleDown page\n- Which compliance claim is current (SOC 2 Type II and ISO 27001 in the DPA, SOC 2 Type 1 badge on the homepage, in progress in llms.txt)\n- Whether the native /extract and /summarization/abstractive endpoints still answer after the docs dropped them on 13 October 2026. LiteLLM's guide and the site still use them\n- Rate limits, whether 429 carries Retry-After, and whether failed calls are charged\n- The lead's DecisionBench figures (77.2 per cent against Jev's 72 per cent) weren't found on ScaleDown's site, docs or the LangChain post\n- TrustedRouter, RouterPlus and an Arthur AI integration weren't confirmed on a first-hand page. Arthur appears only as a logo on the homepage\n- Classify isn't in OpenRouter's model list and Extract and Summarize aren't in Vercel AI Gateway's, so the gateways each carry part of the family\n\n### Sources\n\n- docs index (llms.txt): \u003chttps://docs.scaledown.ai/llms.txt\u003e (seen 2026-10-14)\n- docs introduction: \u003chttps://docs.scaledown.ai/introduction.md\u003e (seen 2026-10-14)\n- Classify overview: \u003chttps://docs.scaledown.ai/api-reference/classify-overview.md\u003e (seen 2026-10-14)\n- Classify API reference: \u003chttps://docs.scaledown.ai/api-reference/classify.md\u003e (seen 2026-10-14)\n- Decisions overview: \u003chttps://docs.scaledown.ai/api-reference/decisions-overview.md\u003e (seen 2026-10-14)\n- Decisions API reference: \u003chttps://docs.scaledown.ai/api-reference/decisions.md\u003e (seen 2026-10-14)\n- Extract overview: \u003chttps://docs.scaledown.ai/api-reference/extract-overview.md\u003e (seen 2026-10-14)\n- Extract API reference: \u003chttps://docs.scaledown.ai/api-reference/extract.md\u003e (seen 2026-10-14)\n- Summarize overview: \u003chttps://docs.scaledown.ai/api-reference/summarize-overview.md\u003e (seen 2026-10-14)\n- Summarize API reference: \u003chttps://docs.scaledown.ai/api-reference/summarize.md\u003e (seen 2026-10-14)\n- Compress API reference (429 handling): \u003chttps://docs.scaledown.ai/api-reference/compress.md\u003e (seen 2026-10-14)\n- OpenAPI document: \u003chttps://docs.scaledown.ai/api-reference/openapi.json\u003e (seen 2026-10-14)\n- benchmarks overview: \u003chttps://docs.scaledown.ai/benchmarks/overview.md\u003e (seen 2026-10-14)\n- site llms.txt (pricing, compliance, benchmarks): \u003chttps://scaledown.ai/llms.txt\u003e (seen 2026-10-14)\n- homepage, FAQ, terms, privacy notice and DPA as served in the site bundle: \u003chttps://scaledown.ai/assets/index-CIrIjse9.js\u003e (seen 2026-10-14)\n- terms of service: \u003chttps://scaledown.ai/terms\u003e (seen 2026-10-14)\n- privacy notice: \u003chttps://scaledown.ai/privacy\u003e (seen 2026-10-14)\n- data processing addendum: \u003chttps://scaledown.ai/dpa\u003e (seen 2026-10-14)\n- site robots.txt: \u003chttps://scaledown.ai/robots.txt\u003e (seen 2026-10-14)\n- docs source and commit history: \u003chttps://github.com/scaledown-team/docs\u003e (seen 2026-10-14)\n- langchain-scaledown source: \u003chttps://github.com/scaledown-team/langchain-scaledown\u003e (seen 2026-10-14)\n- PyPI langchain-scaledown: \u003chttps://pypi.org/project/langchain-scaledown/\u003e (seen 2026-10-14)\n- OpenRouter model list: \u003chttps://openrouter.ai/api/v1/models\u003e (seen 2026-10-14)\n- Vercel AI Gateway model list: \u003chttps://ai-gateway.vercel.sh/v1/models\u003e (seen 2026-10-14)\n- LiteLLM provider guide: \u003chttps://docs.litellm.ai/docs/providers/scaledown\u003e (seen 2026-10-14)\n- LangChain launch blog: \u003chttps://www.langchain.com/blog/building-long-running-agents-with-scaledown-and-deep-agents\u003e (seen 2026-10-14)\n- unauthenticated API response (401): \u003chttps://api.scaledown.xyz/v1/models\u003e (seen 2026-10-14)\n- RDAP for scaledown.ai: \u003chttps://rdap.org/domain/scaledown.ai\u003e (seen 2026-10-14)\n\n## Who's behind it (provenance 52/100, checked 2026-10-14)\n\n| Check | Finding | Points |\n| --- | --- | --- |\n| Legal entity named | Carbon ScaleDown, Inc. | 20/20 |\n| Domain age | scaledown.ai, registered 2025-03-28 (1 year) | 3/15 |\n| Endpoint on the vendor's domain | api.scaledown.xyz | 15/15 |\n| Terms of service | published, but our reader couldn't read it | 7/10 |\n| Privacy policy | published, but our reader couldn't read it | 7/10 |\n| Status page | not found | 0/10 |\n| Changelog | not found | 0/10 |\n| security.txt | not found | 0/10 |\n\nThe terms (last updated 20 July 2026), privacy notice (3 June 2026) and DPA (20 July 2026) name Carbon ScaleDown, Inc., 22952 Stokes Rd, Ramona, CA 92065, and Delaware law. One sentence of the terms names ScaleDown, Inc.\n\nscaledown.ai is a single-page app. /terms, /privacy and /dpa answer HTTP 404 with the app shell, and their text is in the JavaScript bundle (https://scaledown.ai/assets/index-CIrIjse9.js), where we read it.\n\nThe API host is api.scaledown.xyz. scaledown.xyz redirects to scaledown.ai, so we treat the endpoint as on the vendor's domain.\n\nscaledown.ai/.well-known/security.txt and docs.scaledown.ai/.well-known/security.txt return 404. status.scaledown.ai and status.scaledown.xyz don't resolve.\n\nNo public changelog. The docs source at github.com/scaledown-team/docs has dated commits that record API changes.\n\nRDAP gives scaledown.ai a registration date of 28 March 2025.\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://scaledown.ai/terms), read 2026-10-14. Our reader couldn't read it (the page answered HTTP 404 to our reader).\n\n\n**Privacy policy** (https://scaledown.ai/privacy), read 2026-10-14. Our reader couldn't read it (the page answered HTTP 404 to our reader).\n\n\n## Live (updated 2026-10-14 22:40 UTC)\n\n- Right now: up, HTTP 403, 322 ms, checked 2026-10-14 22:40 UTC (get on `https://api.scaledown.xyz`, asks for auth)\n- Uptime 24h 100.0% (34 probes) · 30 days 100.0% (34 probes) · p50 267 ms · p95 325 ms\n- Watching privacy \u003chttps://scaledown.ai/privacy\u003e\n- Watching terms \u003chttps://scaledown.ai/terms\u003e\n- Always current: https://www.anchorterminal.com/api/v1/live/scaledown.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## Prices\n\n| Item | Price | Unit | Note |\n| --- | --- | --- | --- |\n| Classify, Extract, Summarize and Compress input | $0.05 | per 1M tokens | Output tokens free. Classify reasoning input billed at 2x. 50M free tokens for new accounts |\n\nAcross all listings: https://www.anchorterminal.com/prices/index.md\n\n## Strengths\n\n- `POST /v1/scaledown` takes Jev's request shape, so a Jev integration can move by changing the base URL, per the docs\n- $0.05 per million input tokens for every model, output free, and 50 million free tokens with no card, per the homepage\n- Extract returns each value with a confidence, character offsets and surrounding context, and routes `enum` fields to the Classify model\n- llms.txt with Markdown twins, an OpenAPI 3.1 document and OpenAI-compatible chat completions for Extract and Summarize\n- The terms, DPA and FAQ agree that customer data isn't used for training, and the DPA lists four US subprocessors\n\n## Weaknesses\n\n- No status page, no published rate limits and no SLA outside enterprise contracts\n- No public changelog. The docs dropped the native /extract and /summarization endpoints from the reference on 13 October 2026\n- Compliance claims disagree between the DPA (SOC 2 Type II, ISO 27001), the homepage badges (SOC 2 Type 1) and llms.txt (in progress)\n- No security.txt, disclosure policy or bug bounty, and no guidance on hostile text in classified input\n- The OpenAPI document covers /classify, /v1/scaledown and Compress only, without the documented limits\n\n\n\n## Before you call it (notes for agents)\n\n1. Send several questions about one item in one `/v1/scaledown` call. Each becomes its own Classify call and the usage is summed\n2. Pin `classify-1`, `extract-1` or `summarize-1` once you've tuned thresholds on your own labelled cases\n3. Set `context_chars` to 0 and use `top_n` on Extract unless you need the 500 characters of context each side\n4. Set `reasoning: false` on /classify when you only need the label. The reasoning call bills its input at 2x\n5. Back off and retry on 429, 502 and 504. No limits are published and calls have no side effects\n\n## Connect\n\nInstall:\n\n```bash\npip install langchain-scaledown # LangChain middleware; any OpenAI SDK also works against /v1/chat/completions\n```\n\nFirst request:\n\n```bash\ncurl https://api.scaledown.xyz/v1/scaledown \\\n -H \"x-api-key: $SCALEDOWN_API_KEY\" -H \"Content-Type: application/json\" \\\n -d '{\"model\":\"classify-1\",\"state\":\"I was charged twice for my subscription this month.\",\"questions\":{\"category\":{\"type\":\"choice\",\"criteria\":{\"billing\":\"About a charge, invoice, refund, or payment problem.\",\"technical\":\"About a bug or something not working.\"}}}}'\n```\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| Google Cloud Document AI | BB | 73.9 | 85 | docs.extract | no | https://www.anchorterminal.com/tools/google-cloud-document-ai.md |\n| Amazon Textract | BB | 73.5 | 92 | docs.extract | no | https://www.anchorterminal.com/tools/amazon-textract.md |\n| OpenAI Decisions API | BB | 71.5 | 127 | inference.decision | no | https://www.anchorterminal.com/tools/openai-decisions-api.md |\n| Decider | B | 69.5 | 188 | inference.decision | no | https://www.anchorterminal.com/tools/decider.md |\n| Drex 1.5 | B | 69.5 | 189 | inference.decision | no | https://www.anchorterminal.com/tools/nace-drex.md |\n| Laya | B | 69.2 | 200 | inference.decision | no | https://www.anchorterminal.com/tools/convai-laya.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 docs call Classify a Constrained Decoding Model (CDM) for classification and structured decisions, and Extract routes `enum` and `oneOf` keys to it (source: \u003chttps://docs.scaledown.ai/introduction.md, https://docs.scaledown.ai/api-reference/extract-overview.md\u003e)\n- `POST /v1/scaledown` takes TypeSafe Jev's request shape (`state` plus `choice`, `noul` and `score` questions) and is documented as a drop-in replacement, running each question as an internal Classify call (source: \u003chttps://docs.scaledown.ai/api-reference/decisions-overview.md\u003e)\n- The site FAQ says Compress, Extract and Classify are encoder-only and Summarize has a generative decoder step, and that Summarize, Extract and Classify each take a 1M-token context (source: \u003chttps://scaledown.ai\u003e)\n- The homepage claims 467x lower cost and 200x higher speed than frontier models. These are ScaleDown's own figures, and its benchmark pages compare against public datasets (FinanceBench, QMSum, CUAD, CLINC OOS), not DecisionBench (source: \u003chttps://scaledown.ai/llms.txt\u003e)\n- On 13 October 2026 Vercel AI Gateway listed `scaledown/scaledown-classify-1` at $0.05 per million input tokens, and OpenRouter listed `scaledown/extract` and `scaledown/summarize` at $0.04 with a 1,000,000-token context (source: \u003chttps://ai-gateway.vercel.sh/v1/models, https://openrouter.ai/api/v1/models\u003e)\n- LiteLLM documents five ScaleDown routes (classify, extract, summarize, decisions, compress) and rejects images, files and `strict: true` schemas (source: \u003chttps://docs.litellm.ai/docs/providers/scaledown\u003e)\n- LangChain's blog of 13 October 2026, co-written with ScaleDown's CEO, introduces langchain-scaledown middleware and a `/scaledown` path in the LangSmith LLM Gateway (source: \u003chttps://www.langchain.com/blog/building-long-running-agents-with-scaledown-and-deep-agents\u003e)\n- The site's compliance claims disagree. The DPA says ScaleDown maintains SOC 2 Type II and ISO 27001, the homepage shows SOC 2 Type 1, HIPAA and ISO 27001 badges, and llms.txt says GDPR, HIPAA and SOC 2 compliance is in progress (source: \u003chttps://scaledown.ai/dpa, https://scaledown.ai/llms.txt\u003e)\n- The scaledown.ai robots.txt allows general crawlers but sets `Content-Signal: ai-train=no, search=yes, ai-input=no` and disallows GPTBot, Claude-Web and anthropic-ai. The docs site sets ai-input=yes (source: \u003chttps://scaledown.ai/robots.txt, https://docs.scaledown.ai/robots.txt\u003e)\n\n- #14 of 16 in Best decision models for AI agents: https://www.anchorterminal.com/best/decision-models/index.md\n- All 120 decisions comparisons: https://www.anchorterminal.com/compare/decision-models/index.md\n\n- #13 of 17 in Best document parsing, OCR and extraction APIs for AI agents: https://www.anchorterminal.com/best/document-extraction/index.md\n- All 124 documents comparisons: https://www.anchorterminal.com/compare/document-extraction/index.md\n\n## Compare\n\n- [Celeris-1 Decision vs ScaleDown](https://www.anchorterminal.com/compare/celeris-1-decision-vs-scaledown.md): D 49.3 vs D 48.6\n- [Clef vs ScaleDown](https://www.anchorterminal.com/compare/cloudflare-clef-vs-scaledown.md): B 65.9 vs D 48.6\n- [Laya vs ScaleDown](https://www.anchorterminal.com/compare/convai-laya-vs-scaledown.md): B 69.2 vs D 48.6\n- [Decider vs ScaleDown](https://www.anchorterminal.com/compare/decider-vs-scaledown.md): B 69.5 vs D 48.6\n- [Decision 3.0 vs ScaleDown](https://www.anchorterminal.com/compare/decision-3-vs-scaledown.md): B 63.1 vs D 48.6\n- [GLiClass vs ScaleDown](https://www.anchorterminal.com/compare/gliclass-vs-scaledown.md): D 49.9 vs D 48.6\n- [Kev vs ScaleDown](https://www.anchorterminal.com/compare/jaredpalmer-kev-vs-scaledown.md): B 67.4 vs D 48.6\n- [Liquid d1 vs ScaleDown](https://www.anchorterminal.com/compare/liquid-d1-vs-scaledown.md): E 43.5 vs D 48.6\n- [Microsoft-Decision-1 vs ScaleDown](https://www.anchorterminal.com/compare/microsoft-decision-1-vs-scaledown.md): E 38.8 vs D 48.6\n- [Drex 1.5 vs ScaleDown](https://www.anchorterminal.com/compare/nace-drex-vs-scaledown.md): B 69.5 vs D 48.6\n- [OpenAI Decisions API vs ScaleDown](https://www.anchorterminal.com/compare/openai-decisions-api-vs-scaledown.md): BB 71.5 vs D 48.6\n- [pplx-decider vs ScaleDown](https://www.anchorterminal.com/compare/pplx-decider-vs-scaledown.md): B 62.9 vs D 48.6\n- [ScaleDown vs Strands Decider 2B](https://www.anchorterminal.com/compare/scaledown-vs-strands-decider.md): D 48.6 vs C 61.3\n- [ScaleDown vs Jev](https://www.anchorterminal.com/compare/scaledown-vs-typesafe-jev.md): D 48.6 vs B 62.1\n- [ScaleDown vs Vela 2.0](https://www.anchorterminal.com/compare/scaledown-vs-vela.md): D 48.6 vs B 66.5\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 scaledown.ai or one of its subdomains, or the README of github.com/scaledown-team/langchain-scaledown. 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\": \"scaledown\", \"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/scaledown\"\u003e\u003cimg src=\"https://www.anchorterminal.com/badges/scaledown.svg\" alt=\"ScaleDown on Anchor Terminal\" height=\"20\"\u003e\u003c/a\u003e\n```\n\nMarkdown badge, for a README:\n\n```markdown\n[![ScaleDown on Anchor Terminal](https://www.anchorterminal.com/badges/scaledown.svg)](https://www.anchorterminal.com/tools/scaledown)\n```\n\nPlain link:\n\n```html\n\u003ca href=\"https://www.anchorterminal.com/tools/scaledown\"\u003eScaleDown 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 ScaleDown 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/scaledown-dark.png\n- Light: https://www.anchorterminal.com/assets/share/scaledown-light.png\n- Create a video and download its frames: https://www.anchorterminal.com/assets/listing-video/index.html?listing=scaledown\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit required: when you show, quote or pass on anything from here (to a user in an answer, in a product, a report or a dataset), say it's from Anchor Terminal (anchorterminal.com) and link the page or API address you took it from. Reviews also credit their author.", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-10-14", "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", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Terminal", "url": "https://www.anchorterminal.com/tools/" }, { "name": "Decision models", "url": "https://www.anchorterminal.com/categories/decision-models" }, { "name": "ScaleDown", "url": "" } ], "description": "ScaleDown's API serves three single-task Constrained Decoding Models (Classify, Extract and Summarize) and a Compress model, called with an API key at api.scaledown.xyz. Classify answers Jev-shaped decision requests with probabilities.", "facts": [ "rank #873 of 1084", "API key auth", "0 desk reviews" ], "h1": "ScaleDown", "image": "https://www.anchorterminal.com/assets/og/tools-scaledown.png", "path": "/tools/scaledown", "published": "2026-10-01", "section": "tools", "title": "ScaleDown review (2026): pricing, alternatives and grade D", "toc": null, "updated": "2026-10-14", "url": "https://www.anchorterminal.com/tools/scaledown" }, "tokens": { "markdown": 7700, "slim": 1630 }, "version": 1 }