AssemblyAI Speech-to-Text (Universal) by AssemblyAI

# Fix list: AssemblyAI Speech-to-Text (Universal)

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/assemblyai-stt, the October 2026 research run, assessed 1 October 2026. Grade B, 67 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 AssemblyAI Speech-to-Text (Universal): 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. Security & auth, 50 out of 100, up to 8.8 more on the total

Why it scored 50: Model reading of the checklist, with training and retention in place of least-privilege and injection lines. Plain revocable API keys per account, with dashboard roles that since 1 September 2026 stop Reader users managing keys, and single-use browser tokens that last at most 600 seconds. Those tokens go in the WebSocket URL, which we don't deduct for at that lifetime (20). Training is on by default. Paid accounts opt out in the dashboard, free accounts can't (5). Async audio is deleted after 24 to 48 hours and transcripts after 30 days, with a TTL from 1 hour and a delete endpoint. Streaming is zero-retention only once opted out (10). Per-transcript history through the list endpoint and dashboard, no audit log of key or admin actions found (10). SOC 2 Type 2 with a Vanta trust centre (5). No security.txt, disclosure policy, bug bounty or public advisories found (0).

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.

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

Why it scored 40: No x402, MPP or L402 (0). Per-hour prices for every model and add-on published without a login (20). Free tier of up to 185 hours pre-recorded or 333 hours streaming with no card (20). A person has to sign up in a browser to get a key (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. Reliability, 70 out of 100, up to 6 more on the total

Why it scored 70: Statuspage at status.assemblyai.com with component history (20). Twelve incidents between 7 July and 28 September 2026. Two count as major, a US async outage on 16 September that failed about half of async jobs for 75 minutes and a us-east Universal-3.5 Pro streaming fault on 31 July that returned no transcripts for about 2 hours. The rest ran 7 to 67 minutes (5, between the rubric's 10 for one major and 0 for several). Limits published with numbers, 5 or 200+ parallel jobs, 20,000 HTTP requests per 5 minutes, 5 or 100+ new streams a minute (15). Overload handling is partly documented. Jobs queue rather than fail, the docs advise jittered polling and resubmitting after a server error, but the HTTP limit answers 403 rather than 429 and no Retry-After is documented (10). A 99.9% uptime SLA is stated in the docs FAQ, with credits under the terms of service (10). Async, streaming and Sync APIs are 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.

## 4. Agent ergonomics, 80 out of 100, up to 3.3 more on the total

Why it scored 80: API reading of the checklist. Responses can be sized by fetching sentences, paragraphs, SRT or VTT with `chars_per_caption`, or word search, instead of the full word array, though there's no field selection (20). `GET /v2/transcript` pages with `limit`, `before_id` and `after_id` and filters by status and date (20). Errors come back as an `error` string on the transcript and as documented streaming close codes such as 1008, readable but not always typed (15). No idempotency key, and a resubmitted job is a new billed job. The retry guide covers server errors only (10). One required parameter and official SDKs in Python and TypeScript among others (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.

## 5. Transparency & trust, 78 out of 100, up to 1.9 more on the total

Made of editorial 70, provenance 86.

Why it scored 78: Closed service with clear terms (15). Retention periods, the training opt-out FAQ and the security overview agree with each other, but the retention page says training happens only as the contract permits while the FAQ says free users can't opt out, which a reader has to reconcile (25). Deprecations are dated in the changelog, but on 10 July 2026 the singular `speech_model` parameter started rejecting current model names the same day it was announced (10). Subprocessor list at /legal/subprocessors and US and EU endpoints (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: assemblyai.com, registered 2016-12-24 (9 years) (11 of 15)
- security.txt: not found (0 of 10)

## 6. Maintenance & community, 80 out of 100, up to 1.8 more on the total

Why it scored 80: Dated changelog entries on 3 September (speaker confidence scores) and 8 September 2026 for speech-to-text, and Python SDK 1.5.5 on 16 September (30). More than 40 dated entries in the last 90 days (20). Public changelog and support address. The Python SDK has 8 open issues, the newest from December 2025 and the oldest from March 2024, so replies are slow (10). Official SDKs current, Python 1.5.5 supports 3.8 to 3.14 (15). CI on the SDK repositories not confirmed (5).

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.

## 7. Schema & documentation, 95 out of 100, up to 0.8 more on the total

Why it scored 95: OpenAPI 3.1 at /docs/openapi.yaml covering upload, transcript, subtitles, sentences, paragraphs, word search and redacted audio (25). llms.txt at /docs/llms.txt (10). Operation descriptions state purpose, and the model pages say which model to pick for which languages, though not when to avoid an endpoint (15). Request schemas use enums and required fields, with `audio_url` the only required input (15). 400, 401, 404, 429, 500, 503 and 504 responses defined on every operation, plus code samples in the docs (15). Versioned `/v2` paths and a dated public changelog (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.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-07-10, the async `speech_model` parameter began returning 400 for current model names and silently routing legacy names to the default model, and `universal-3-pro` was blocked for new accounts and accounts inactive for 7 days, all announced in the changelog the same day. We found no earlier notice (https://www.assemblyai.com/changelog). Documented, so the minimum deduction.

## 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.

- Whether AssemblyAI gave notice of the 10 July 2026 `speech_model` change before the changelog entry that day
- Whether the 99.9% SLA applies to self-serve paid accounts or only to contracts
- Whether the SDK repositories run CI on the default branch

## Weaknesses

- Two outages of an hour or more in the last 90 days, on 31 July and 16 September 2026
- Free accounts can't opt out of model training
- The HTTP rate limit returns 403 with no Retry-After
- Streaming bills the time the socket is open, up to a 3-hour auto-close
- No security.txt or published disclosure policy

## 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.

- Send `{"type":"Terminate"}` to close every stream, or billing runs to the 3-hour auto-close
- Use `speech_models` (plural). The singular `speech_model` now returns 400 for current model names
- Treat a 403 on polling as the rate limit and back off with jitter, or use webhooks
- Fetch `/sentences` or `/paragraphs` instead of the full transcript when you only need text
- Opt out in Data Controls on a paid account before sending customer audio

## What the review panel asked for

- Shorten socket auto-close
- Return 429 with Retry-After
- Clarify who gets the SLA

## 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.