Cohere Embed and Rerank by Cohere

# Fix list: Cohere Embed and Rerank

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/cohere-embed, the October 2026 research run, assessed 1 October 2026. Grade BB, 72.5 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 Cohere Embed and Rerank: 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: Trial and production API keys, revocable from the dashboard, with no scopes or expiry that we found (20). The same key reaches datasets, fine-tuning, connectors and embed jobs, which include delete operations, with no way to limit it to embed and rerank (10 of 20). Returns vectors and scores, and rerank returns the caller's own documents (10). No per-key log or audit trail found in the docs we read (0 of 15). SOC 2 Type II and a bug bounty on the security page, a Trust Center for documents, but no security.txt on cohere.com and no public advisories found (10 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.

## 2. Payments & pricing, 35 out of 100, up to 8.1 more on the total

Why it scored 35: No x402, MPP or L402 (0). Embed 5 prices are public, $0.12 per million text tokens on Pro and $0.08 on Fast, $0.40 per million image tokens, but the rerank rate per 1,000 searches didn't render on the pricing page, so half for the reranker (15 of 20). Trial keys are created at signup with no card, free and capped at 1,000 calls a month (20). A person signs up in a browser (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, 83 out of 100, up to 3.4 more on the total

Why it scored 83: status.cohere.com (incident.io) lists every embed and rerank model as its own component, embed-v4.0, the v3 embed models, rerank-v4.0-pro, rerank-v4.0-fast and rerank-v3.5 (20). All show 100 per cent from July to October 2026 with no incidents posted. Embed 5 isn't a component yet, a day after launch (30). Published limits, embed 2,000 text inputs a minute on trial and production keys, rerank 10 requests a minute on trial and 1,000 on production, 1,000 calls a month on trial keys (15). The errors page lists 429 with a note to retry with backoff, but no Retry-After header or timing, and 500s are sent to support (8 of 15). No SLA found for any self-serve tier (0). Rerank 4 and Embed 5 are on the production API, not marked preview (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. Transparency & trust, 69 out of 100, up to 2.7 more on the total

Made of editorial 47, provenance 90.

Why it scored 69: Closed service under published terms, SDKs MIT (15). The terms let Cohere use customer data to improve the service, including by sharing API data with third parties, the model training notice says training happens only with permission, and the security page says you can opt out of training at any time, which implies it's on by default. Three statements that don't agree (12 of 30). Deprecations page defines the deprecated state and lists models with shutdown dates, embed v2 retired on 4 April 2026, but states no minimum notice period (15 of 20). A Trust Center holds compliance documents and the privacy policy names Toronto, but we found no public subprocessor list or data-location statement for the API (5 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):

- security.txt: not found (0 of 10)

## 5. Agent ergonomics, 87 out of 100, up to 2.1 more on the total

Why it scored 87: output_dimension from 256 to 2048, output types float, int8, uint8, binary, ubinary and base64, and top_n on rerank (25). truncate NONE, START or END and max_tokens_per_doc on rerank, but only 96 inputs a call (17 of 20). Status codes listed with meanings, and 400s point at the request, though the guidance on what to change is thin (14 of 20). Embedding and rerank calls are stateless and the errors page says to retry 429s with backoff (18 of 20). input_type is a second required field that trips first calls, official SDKs in Python, TypeScript, Go and Java (13 of 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, 92 out of 100, up to 1.3 more on the total

Why it scored 92: cohere-openapi.yaml is public in cohere-ai/cohere-developer-experience (25). llms.txt at docs.cohere.com (10). The reference explains each input_type and what it's for, and the rerank reference says when to set max_tokens_per_doc and how many documents to send (15 of 20). model and input_type required, input_type, embedding_types and truncate are enums, 96 inputs a call is stated (15). Examples on every reference page and status codes 400 to 504 listed on the embed reference, but no error bodies there (12 of 15). Versioned v2 API and a dated 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.

## 7. Maintenance & community, 87 out of 100, up to 1.1 more on the total

Why it scored 87: Embed 5 Pro and Fast shipped on 30 September 2026 (30). Four dated changelog entries since 3 July 2026, embed-v5, north-small-translate, parse and transcribe-arabic (20). cohere-python has only 4 open issues and 20 open pull requests, so issues do get closed, but an embed bug opened on 15 August 2026 (batched embeddings drop types absent from the first response) shows no reply (15 of 25). Official SDKs in four languages, and the Python release list tops out at 7.1.0, which adds the /parse endpoint that launched on 27 August 2026 (15). CI on GitHub Actions, but PyPI showed 6.1.0 to our fetch while GitHub lists 7.1.0, so the published version is unclear (7 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.

- Which statement governs training on API data. The terms allow improvement use and sharing with third parties, the training notice requires permission, and the security page describes an opt-out.
- The per-search price for Rerank 4 on Cohere's own API.
- Which cohere-python version is current on PyPI. Our fetch showed 6.1.0, GitHub releases list 7.1.0.
- Whether Embed 5 takes PDFs directly or only through the new parse endpoint. The launch post names text, images and fused text and image.

## Weaknesses

- Terms, training notice and security page disagree on whether API data trains models or goes to third parties
- The per-search rerank price didn't render on the pricing page
- 96 inputs a call, and input_type is required
- One unscoped key reaches every Cohere endpoint, including delete operations
- No security.txt, no published subprocessor list found, no SLA

## 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 input_type on every embed call, search_document when indexing and search_query when querying. The endpoint rejects a call without it
- Batch 96 inputs a call, the maximum, stay under 2,000 inputs a minute, and check every batch returns every embedding type you asked for (an open SDK bug drops types missing from the first response)
- Budget rerank by searches. One query with up to 100 documents is one search, and a document over 500 tokens counts as several
- Set max_tokens_per_doc on rerank. The default of 4,096 truncates long documents even on the 32K models
- Ask for int8 or binary embedding_types and a smaller output_dimension before scaling the vector store

## What the review panel asked for

- Publish rerank per-search rate
- Show an error body for each status
- Name Retry-After on 429

## 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.