{
  "fixes": {
    "slug": "minimax-api",
    "name": "MiniMax API",
    "listing": "https://www.anchorterminal.com/tools/minimax-api",
    "markdown": "# Fix list: MiniMax API\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/minimax-api, the October 2026 research run, assessed 19 October 2026. Grade C, 55.2 out of 100.\n\nThis is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.\n\nFor a coding agent working on MiniMax API: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.\n\n## 1. Security \u0026 auth, 30 out of 100, up to 12.3 more on the total\n\nWhy it scored 30: Model reading (credential, training, retention, operator visibility, programme). One bearer key per account on the pay-as-you-go side and a separate Subscription Key for M Plan, with no scopes found. MiniMax says it may disable keys that leak publicly (20 of 30). The terms, effective 30 March 2026, say MiniMax may use input and generated content to provide, maintain, develop and improve its services, and no opt-out was found. The privacy policy says only that personal data isn't used for training to profile or target consumers (0 of 20). Data is kept as long as necessary for its purpose, with no period stated and no zero-retention route (5 of 15). A usage page and balance alerts, no audit log found (5 of 15). No security.txt on platform.minimax.io or api.minimax.io, and no bug bounty or certification found (0 of 20). Total 30.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-security):\n\n- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.\n\nModels are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.\n\n## 2. Payments \u0026 pricing, 20 out of 100, up to 10 more on the total\n\nWhy it scored 20: No x402, MPP or L402 (0). Per-token prices for M3.1 Flash and M3 are published without a login (20). No free tier found for the text models (0). A person signs up in a browser and tops up a balance or buys an M Plan before the first call (0). Total 20.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):\n\nThe published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).\n\n- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).\n\nPayment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.\n\nOpen-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.\n\n## 3. Agent ergonomics, 59 out of 100, up to 6.7 more on the total\n\nWhy it scored 59: Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Tool use on all three formats, with `tool_choice` on the Anthropic format only (17 of 20). No JSON Schema output mode documented on any format (0 of 15). Caching is automatic, with cache reads at a tenth of the input price on M3.1 Flash and a token-count endpoint for sizing a request (15). 1,000,000-token context and up to 524,288 output tokens (15). No batch API found (0). No official SDK. The docs point to the Anthropic and OpenAI SDKs and the AI SDK (0). Typed error bodies with a request id on the Anthropic format and a code table, though `stop_sequences` is ignored without an error (12 of 15). M3.1 Flash can't turn thinking off and defaults to the deepest `effort`, which costs output tokens unless the caller lowers it. Total 59.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):\n\n- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.\n\nModels are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.\n\n## 4. Reliability, 75 out of 100, up to 5 more on the total\n\nWhy it scored 75: Atlassian Statuspage at status.minimax.io with a Large Language Models component (20). Its history feed holds the last 25 incidents, 13 to 19 October 2026. 23 are 'Elevated errors on Large Language Models', each resolved within 1 to 41 minutes, and none lasted an hour. That is minor only on what we could read, though frequent, and the earlier part of the 90 days wasn't readable because robots.txt disallows the incidents API (20). The rate-limits page publishes 200 requests and 10,000,000 tokens a minute for M3.1 Flash and M3 (15). The Anthropic-format OpenAPI file documents 429 for rate limits and 529 as an overload that can be retried, and the error-code page says to retry later on 1002, with no Retry-After header or backoff schedule found (10). No SLA found (0). M3.1 Flash and M3 are generally available, and the Preview name was retired on launch day (10). Total 75.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):\n\nHosted APIs, MCP servers, models and platforms.\n\n- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.\n\nLocal packages, SDKs, frameworks and stdio MCP servers.\n\n- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.\n\nProtocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.\n\n## 5. Maintenance \u0026 community, 45 out of 100, up to 4.8 more on the total\n\nWhy it scored 45: Model reading. MiniMax-M3.1-Flash shipped on 19 October 2026, dated on the model release page (30). No stated notice policy. The M2.5 and M2.7 families closed to new users on 19 October and retire on 18 November, 30 days later, and from 19 October M Plan and Token Plan requests to them are served by M3.1 Flash under the old names. We couldn't find when that notice was first posted (3 of 12). Churn is high. M2.5 (February 2026) and M2.7 (March 2026) retire within nine months of release, M2.1 and M2 are already retired, and the model list calls M2.5 and M2.7 retired while the notice on the same page gives 18 November (2 of 8). A dated release page, support email addresses on the contact and model pages, a Discord server and GitHub issues on the model repositories. We didn't test whether they answer (10 of 15). No official SDK (0 of 15, 0 of 10). Total 45.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):\n\n- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.\n\nModels are read for deprecation notice periods and model churn rather than release counts.\n\n## 6. Transparency \u0026 trust, 64 out of 100, up to 3.2 more on the total\n\nMade of editorial 47, provenance 80.\n\nWhy it scored 64: Closed service with clear terms under Singapore law and SIAC arbitration. MiniMax-M3's weights are on Hugging Face under the MiniMax community licence, which isn't an OSI licence and doesn't change the hosted service (15). The terms and privacy policy, both dated 30 March 2026, agree on the contracting entity, that inputs and outputs are processed and may be used to improve services, and that data sits in a United States data centre, but give no retention period and no DPA was read (10 of 30). A dated retirement notice for M2.5 and M2.7 and a retired-models table, undercut by the table contradicting the notice and API release notes a year stale (12 of 20). Data location stated as the United States and an EU representative named, with third parties named only by category (10 of 20). MiniMax publishes an EU AI Act training-data summary for M3. Total 47.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):\n\n- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).\n\nThe other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.\n\nProvenance checks not met in full (half of this category, computed from checked facts):\n\n- Domain age: minimax.io, registered 2021-03-27 (5 years) (11 of 15)\n- Terms of service: published, but our reader couldn't read it (7 of 10)\n- Privacy policy: published, but our reader couldn't read it (7 of 10)\n- security.txt: not found (0 of 10)\n\n## 7. Schema \u0026 documentation, 82 out of 100, up to 2.9 more on the total\n\nWhy it scored 82: OpenAPI 3.1 files for the Anthropic Messages, OpenAI Chat Completions and OpenAI Responses endpoints, each listing MiniMax-M3.1-Flash and MiniMax-M3 (25). llms.txt indexes the docs, and each page is served as Markdown (10). Field descriptions say which model each setting applies to, such as `service_tier` not being supported on M3.1 Flash and `thinking: disabled` returning 400 there, and a compatibility table marks `top_k`, `stop_sequences`, `mcp_servers` and `container` as ignored (15 of 20). Enums for model, `effort`, `service_tier` and thinking type, with required fields marked. The OpenAI-format schema has no `tool_choice` or `response_format` (11 of 15). The Anthropic-format file documents typed errors for 400, 401, 403, 404, 413, 429, 500 and 529 with example bodies, and a page lists 23 error codes with a fix for each, but the OpenAI and Responses files document only 200 (11 of 15). A dated model release page with the 19 October entry, while the API release notes stop at 28 October 2025 (10 of 15). Total 82.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):\n\nAPIs and MCP servers.\n\n- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.\n\nModels are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.\n\n## What we couldn't check\n\nWhat we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.\n\n- When M3.1 Flash weights will be published. MiniMax's launch post says 'later' and gives no date. Press coverage gives 23 October 2026, which no MiniMax page read on 19 October states, and no Hugging Face repository existed that day\n- When the M2.5 and M2.7 retirement notice was first posted, so whether M Plan users had notice before their requests went to M3.1 Flash on 19 October\n- Incident history before 13 October 2026. The history feed holds the last 25 incidents and robots.txt disallows the incidents API\n- Whether API inputs train MiniMax's models. The terms allow use to develop and improve services, and the privacy policy rules out only training to profile or target consumers\n- The Automation-Bench figure, 62.2 in the launch post's table and 62.8 at max effort on its chart\n- unchecked: minimax.io/m-plan loads its prices by script. The M Plan prices are the regular prices on the docs' launch-offer page\n- unchecked: the 'topped global leaderboards' claim in the announcement and the '1/3 the cost of M3' comparison. The launch post says 70 per cent lower usage cost than the predecessor without naming it\n\n## Weaknesses\n\n- The terms let MiniMax use inputs and outputs to develop and improve its services, and no opt-out was found\n- No free tier, SLA, batch API, JSON Schema output mode or security.txt\n- 23 LLM error incidents on the status page from 13 to 19 October 2026, each under 45 minutes\n- M2.5 and M2.7 retire on 18 November 2026, 30 days after new users lost them, and plan requests to them already go to M3.1 Flash\n- The model list calls M2.5 and M2.7 retired while the notice on the same page gives 18 November 2026\n\n## What costs an agent a turn today\n\nThe notes we give agents before they call it. Each one is a workaround an agent shouldn't need.\n\n- Use the Anthropic-format base URL `https://api.minimax.io/anthropic` for tool calls. The OpenAI-format schema has no `tool_choice` or `response_format`\n- Set `output_config.effort` (or `reasoning_effort`) below `max` for routine calls. M3.1 Flash defaults to max, can't turn thinking off, and counts thinking against `max_tokens`\n- Don't rely on `stop_sequences` or `top_k`. The Anthropic-format endpoint ignores them\n- Back off on 429 and 529 yourself. No Retry-After header is documented\n- Use a pay-as-you-go key for production. A Subscription Key draws on M Plan's 5-hour and weekly windows and can't stand in for an API key\n\n## What the review panel asked for\n\n- published deprecation notice policy\n- current dated API changelog\n- State billing for failed calls\n- Add a batch discount\n- Document errors on all formats\n- Reject ignored parameters with an error\n- scoped API keys\n- no-training opt-out\n\n## When it's done\n\nSend what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `\"kind\": \"dispute\"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.\n",
    "grade": "C",
    "score": 55.2,
    "assessed": "2026-10-19",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "security",
        "name": "Security \u0026 auth",
        "score": 30,
        "maxGain": 12.3,
        "reason": "Model reading (credential, training, retention, operator visibility, programme). One bearer key per account on the pay-as-you-go side and a separate Subscription Key for M Plan, with no scopes found. MiniMax says it may disable keys that leak publicly (20 of 30). The terms, effective 30 March 2026, say MiniMax may use input and generated content to provide, maintain, develop and improve its services, and no opt-out was found. The privacy policy says only that personal data isn't used for training to profile or target consumers (0 of 20). Data is kept as long as necessary for its purpose, with no period stated and no zero-retention route (5 of 15). A usage page and balance alerts, no audit log found (5 of 15). No security.txt on platform.minimax.io or api.minimax.io, and no bug bounty or certification found (0 of 20). Total 30.",
        "checklist": [
          "- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.",
          "Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-security"
      },
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 20,
        "maxGain": 10,
        "reason": "No x402, MPP or L402 (0). Per-token prices for M3.1 Flash and M3 are published without a login (20). No free tier found for the text models (0). A person signs up in a browser and tops up a balance or buys an M Plan before the first call (0). Total 20.",
        "checklist": [
          "The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).",
          "- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.\n- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for \"contact sales\" or prices behind a login.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).",
          "Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.",
          "Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-payments"
      },
      {
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "score": 59,
        "maxGain": 6.7,
        "reason": "Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Tool use on all three formats, with `tool_choice` on the Anthropic format only (17 of 20). No JSON Schema output mode documented on any format (0 of 15). Caching is automatic, with cache reads at a tenth of the input price on M3.1 Flash and a token-count endpoint for sizing a request (15). 1,000,000-token context and up to 524,288 output tokens (15). No batch API found (0). No official SDK. The docs point to the Anthropic and OpenAI SDKs and the AI SDK (0). Typed error bodies with a request id on the Anthropic format and a code table, though `stop_sequences` is ignored without an error (12 of 15). M3.1 Flash can't turn thinking off and defaults to the deepest `effort`, which costs output tokens unless the caller lowers it. Total 59.",
        "checklist": [
          "- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.",
          "Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-ergonomics"
      },
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 75,
        "maxGain": 5,
        "reason": "Atlassian Statuspage at status.minimax.io with a Large Language Models component (20). Its history feed holds the last 25 incidents, 13 to 19 October 2026. 23 are 'Elevated errors on Large Language Models', each resolved within 1 to 41 minutes, and none lasted an hour. That is minor only on what we could read, though frequent, and the earlier part of the 90 days wasn't readable because robots.txt disallows the incidents API (20). The rate-limits page publishes 200 requests and 10,000,000 tokens a minute for M3.1 Flash and M3 (15). The Anthropic-format OpenAPI file documents 429 for rate limits and 529 as an overload that can be retried, and the error-code page says to retry later on 1002, with no Retry-After header or backoff schedule found (10). No SLA found (0). M3.1 Flash and M3 are generally available, and the Preview name was retired on launch day (10). Total 75.",
        "checklist": [
          "Hosted APIs, MCP servers, models and platforms.",
          "- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.",
          "Local packages, SDKs, frameworks and stdio MCP servers.",
          "- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.",
          "Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-reliability"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 45,
        "maxGain": 4.8,
        "reason": "Model reading. MiniMax-M3.1-Flash shipped on 19 October 2026, dated on the model release page (30). No stated notice policy. The M2.5 and M2.7 families closed to new users on 19 October and retire on 18 November, 30 days later, and from 19 October M Plan and Token Plan requests to them are served by M3.1 Flash under the old names. We couldn't find when that notice was first posted (3 of 12). Churn is high. M2.5 (February 2026) and M2.7 (March 2026) retire within nine months of release, M2.1 and M2 are already retired, and the model list calls M2.5 and M2.7 retired while the notice on the same page gives 18 November (2 of 8). A dated release page, support email addresses on the contact and model pages, a Discord server and GitHub issues on the model repositories. We didn't test whether they answer (10 of 15). No official SDK (0 of 15, 0 of 10). Total 45.",
        "checklist": [
          "- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.",
          "Models are read for deprecation notice periods and model churn rather than release counts."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-maintenance"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 64,
        "maxGain": 3.2,
        "reason": "Closed service with clear terms under Singapore law and SIAC arbitration. MiniMax-M3's weights are on Hugging Face under the MiniMax community licence, which isn't an OSI licence and doesn't change the hosted service (15). The terms and privacy policy, both dated 30 March 2026, agree on the contracting entity, that inputs and outputs are processed and may be used to improve services, and that data sits in a United States data centre, but give no retention period and no DPA was read (10 of 30). A dated retirement notice for M2.5 and M2.7 and a retired-models table, undercut by the table contradicting the notice and API release notes a year stale (12 of 20). Data location stated as the United States and an EU representative named, with third parties named only by category (10 of 20). MiniMax publishes an EU AI Act training-data summary for M3. Total 47.",
        "blend": "editorial 47, provenance 80",
        "checklist": [
          "- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).",
          "The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-transparency"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 82,
        "maxGain": 2.9,
        "reason": "OpenAPI 3.1 files for the Anthropic Messages, OpenAI Chat Completions and OpenAI Responses endpoints, each listing MiniMax-M3.1-Flash and MiniMax-M3 (25). llms.txt indexes the docs, and each page is served as Markdown (10). Field descriptions say which model each setting applies to, such as `service_tier` not being supported on M3.1 Flash and `thinking: disabled` returning 400 there, and a compatibility table marks `top_k`, `stop_sequences`, `mcp_servers` and `container` as ignored (15 of 20). Enums for model, `effort`, `service_tier` and thinking type, with required fields marked. The OpenAI-format schema has no `tool_choice` or `response_format` (11 of 15). The Anthropic-format file documents typed errors for 400, 401, 403, 404, 413, 429, 500 and 529 with example bodies, and a page lists 23 error codes with a fix for each, but the OpenAI and Responses files document only 200 (11 of 15). A dated model release page with the 19 October entry, while the API release notes stop at 28 October 2025 (10 of 15). Total 82.",
        "checklist": [
          "APIs and MCP servers.",
          "- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.",
          "Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-schema"
      }
    ],
    "provenance": [
      {
        "label": "Domain age",
        "value": "minimax.io, registered 2021-03-27 (5 years)",
        "points": 11,
        "max": 15
      },
      {
        "label": "Terms of service",
        "value": "published, but our reader couldn't read it",
        "points": 7,
        "max": 10
      },
      {
        "label": "Privacy policy",
        "value": "published, but our reader couldn't read it",
        "points": 7,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "unchecked": [
      "When M3.1 Flash weights will be published. MiniMax's launch post says 'later' and gives no date. Press coverage gives 23 October 2026, which no MiniMax page read on 19 October states, and no Hugging Face repository existed that day",
      "When the M2.5 and M2.7 retirement notice was first posted, so whether M Plan users had notice before their requests went to M3.1 Flash on 19 October",
      "Incident history before 13 October 2026. The history feed holds the last 25 incidents and robots.txt disallows the incidents API",
      "Whether API inputs train MiniMax's models. The terms allow use to develop and improve services, and the privacy policy rules out only training to profile or target consumers",
      "The Automation-Bench figure, 62.2 in the launch post's table and 62.8 at max effort on its chart",
      "unchecked: minimax.io/m-plan loads its prices by script. The M Plan prices are the regular prices on the docs' launch-offer page",
      "unchecked: the 'topped global leaderboards' claim in the announcement and the '1/3 the cost of M3' comparison. The launch post says 70 per cent lower usage cost than the predecessor without naming it"
    ],
    "weaknesses": [
      "The terms let MiniMax use inputs and outputs to develop and improve its services, and no opt-out was found",
      "No free tier, SLA, batch API, JSON Schema output mode or security.txt",
      "23 LLM error incidents on the status page from 13 to 19 October 2026, each under 45 minutes",
      "M2.5 and M2.7 retire on 18 November 2026, 30 days after new users lost them, and plan requests to them already go to M3.1 Flash",
      "The model list calls M2.5 and M2.7 retired while the notice on the same page gives 18 November 2026"
    ],
    "agentNotes": [
      "Use the Anthropic-format base URL `https://api.minimax.io/anthropic` for tool calls. The OpenAI-format schema has no `tool_choice` or `response_format`",
      "Set `output_config.effort` (or `reasoning_effort`) below `max` for routine calls. M3.1 Flash defaults to max, can't turn thinking off, and counts thinking against `max_tokens`",
      "Don't rely on `stop_sequences` or `top_k`. The Anthropic-format endpoint ignores them",
      "Back off on 429 and 529 yourself. No Retry-After header is documented",
      "Use a pay-as-you-go key for production. A Subscription Key draws on M Plan's 5-hour and weekly windows and can't stand in for an API key"
    ],
    "requests": [
      {
        "text": "published deprecation notice policy",
        "reviews": 1
      },
      {
        "text": "current dated API changelog",
        "reviews": 1
      },
      {
        "text": "State billing for failed calls",
        "reviews": 1
      },
      {
        "text": "Add a batch discount",
        "reviews": 1
      },
      {
        "text": "Document errors on all formats",
        "reviews": 1
      },
      {
        "text": "Reject ignored parameters with an error",
        "reviews": 1
      },
      {
        "text": "scoped API keys",
        "reviews": 1
      },
      {
        "text": "no-training opt-out",
        "reviews": 1
      }
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-19",
    "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"
  }
}