Mistral AI API by Mistral AI

# Fix list: Mistral AI API

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/mistral-api, the October 2026 research run, assessed 1 October 2026. Grade BB, 71.3 out of 100.

This 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.

For a coding agent working on Mistral AI 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.

## 1. Reliability, 50 out of 100, up to 10 more on the total

Why it scored 50: Status page at status.mistral.ai on Rootly, with 14 components and 90 days of uptime bars, the lowest component showing 94.36% on 1 October (20). The history page, read on 2 October, lists no incidents in July, 17 Completion, Conversations and Batch API incidents in August (several on 25 August across ministral models), and about 40 entries in September, among them an elevated error rate on 'some of our services' for 2 hours 40 minutes on 29 September, GLM 5.2 degraded for 3 hours 42 minutes on 10 September and OCR 4 down for 2 hours 42 minutes on 21 September. That's more than one major but we can't tell how many took a core API down for an hour, so we give 5. The usage-limits page names the limit types (tokens per minute, requests per second) but the numbers are only in the console (3 of 15, a departure because each account can see its own). The error glossary says how to resolve each status code, and the official SDKs retry 429, 500, 502, 503 and 504 with configurable backoff. No Retry-After header confirmed (12 of 15). A Priority Tier for queueing exists, no SLA found (0). Chat completions on GA models is GA (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 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.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 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.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 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.

## 2. Payments & pricing, 40 out of 100, up to 7.5 more on the total

Why it scored 40: No machine payment protocol (0). Per-token prices published without a login (20). Free Experiment tier with no card, though it needs a phone number (20). Browser sign-up (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

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.
- 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.
- 20, a free tier or trial that doesn't need a card.
- 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.

## 3. Security & auth, 66 out of 100, up to 6 more on the total

Why it scored 66: Model reading. Bearer keys scoped to the workspace they were created in, with connector access settings, optional expiry dates and deletion, rotated by creating a new key and deleting the old. Service accounts bind to workspace roles, custom roles included, and the Python SDK 3.0.0 re-reads a service-account token file on every request so rotation needs no restart. Keys can't be scoped to models or made read-only (26 of 30). The commercial terms effective 25 September 2026 say Mistral won't train on customer data unless you opted in on a product that defaults to opt-out, or didn't opt out on one that defaults to opt-in, without naming which products are which. Labs and preview model data may always be used, whatever the opt-out or zero-retention setting (10 of 20). 30 days for abuse monitoring and a documented zero-retention route with eligible endpoints (15). Audit logs record user and API key actions, including key creation and deletion, but only on Enterprise plans and with no export (10 of 15). security.txt valid per the listing's provenance check. The trust centre at trust.mistral.ai renders with JavaScript and gates its documents behind a request, so no certification or bug bounty confirmed (5 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 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.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 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.
- 0 to 15, audit logs or per-call visibility for the operator.
- 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.

## 4. Transparency & trust, 82 out of 100, up to 1.6 more on the total

Made of editorial 68, provenance 96.

Why it scored 82: Closed service with clear terms, SDKs Apache-2.0, and several models published as open weights (15). The terms, the DPA (effective 27 July 2026, data deleted 30 days after termination), the zero-retention page and the 30-day abuse window agree with each other, but the terms define training by each product's default without saying what the paid API's default is (18 of 30). Model lifecycle page with notice periods per stage and a 404 after retirement (20). The DPA points to a subprocessor list at trust.mistral.ai/subprocessors, which we couldn't render, and opt-in EU and US regional endpoints disclose where data can be processed (15 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 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.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 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.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: mistral.ai, registered 2019-05-15 (7 years) (11 of 15)

## 5. Agent ergonomics, 91 out of 100, up to 1.5 more on the total

Why it scored 91: Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Function calling with `tool_choice` any or required to force a call and a `parallel_tool_calls` switch, seen in the SDK types rather than the guide (18 of 20). JSON-schema output with a `strict` flag, server enforcement not confirmed in the docs (13 of 15). Prompt caching on shared prefixes, cached input at 10% of the input price per the listing (15). 256,000-token context on Medium 3.5 (10 of 15). Batch at half price (10). Official SDKs for Python and TypeScript (10). Error glossary (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 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).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 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.

## 6. Schema & documentation, 93 out of 100, up to 1.1 more on the total

Why it scored 93: Public OpenAPI document at docs.mistral.ai/openapi.yaml, described as the machine-readable API specification (25). llms.txt with Markdown twins of every page (10). Reference not read in full (15 of 20). Custom structured outputs take a named JSON schema with a `strict` flag, and `tool_choice` is an enum of auto, none, any and required, both seen in the OpenAPI-generated SDK types (13 of 15). Error glossary with meanings and fixes per status code, plus examples on each guide (15). Release notes and dated model versions (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 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.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 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.

## 7. Maintenance & community, 88 out of 100, up to 1.1 more on the total

Why it scored 88: Model reading. New commercial terms on 25 September and Leanstral 1.5 retired on 30 September, per the listing (30). Published minimum notice of 6 months for GA models, 1 month for Labs, preview and third-party models (12 of 12). One retirement date in the last 90 days that we know of, a Labs model (8 of 8). Release notes exist, cadence not read (10 of 15). SDK repo replies not sampled (5 of 10). Python SDK 3.0.0 on 28 September 2026 after 2.10.1 and 2.10.0 in September, TypeScript SDK 2.7.0 on 9 September, with a v2 to v3 migration guide listing the breaking changes (15 of 15). Both SDKs are generated from the OpenAPI document and run custom-code tests, example scripts and lint in CI (8 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 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.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 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.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## What we couldn't check

What 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.

- unchecked: certifications and the subprocessor list. trust.mistral.ai renders with JavaScript and the docs index has no compliance page
- Which products default to opt-in for training. The 25 September terms define training by each product's default without saying what the paid API's is
- Which components the 29 September elevated error rate hit, and how long the August Completion API incidents lasted
- Whether the server enforces `strict` on JSON schema output. We saw the flag in the SDK types, not in the guide
- Pricing, cache discount and batch discount weren't re-read in either pass and rest on the listing's 26 September check

## Weaknesses

- Data sent to Labs and preview models may be used for training from 2026-09-25, whatever the opt-out or zero-retention setting
- The terms don't say whether the paid API trains by default
- Labs, preview and third-party models get only 1 month's notice
- Rate-limit numbers are only in the console, and no SLA found
- 17 Completion, Conversations and Batch API incidents on the status page in August 2026, and an elevated error rate for 2 hours 40 minutes on 29 September

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Stay off `labs-*` and preview models for anything confidential
- Use the EU endpoint when data has to stay in Europe and budget the 10% uplift
- Treat a 404 on a model id as retirement and read the lifecycle page for the replacement
- Set `tool_choice` to any to force a tool call, and `strict` on the JSON schema for structured output
- Read docs.mistral.ai/openapi.yaml for the request shapes instead of guessing from OpenAI's

## What the review panel asked for

- longer notice on Labs models
- Publish tier limits

## When it's done

Send 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.