Voyage AI embeddings and rerankers by Voyage AI (MongoDB)

# Fix list: Voyage AI embeddings and rerankers

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/voyage-ai, the October 2026 research run, assessed 1 October 2026. Grade C, 59 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 Voyage AI embeddings and rerankers: 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, 45 out of 100, up to 11 more on the total

Why it scored 45: status.voyageai.com timed out on 30 September, 1 October and twice on 2 October 2026, and neither the site footer nor the docs index links a status page, so none is credited (0). No readable incident history (5 of 30). Rate limits published by tier, tier 1 once a payment method is added (2,000 requests a minute and 8 million tokens a minute on voyage-3.5, for example), double at $100 paid and treble at $1,000 (15). The rate-limit guide recommends bigger batches, pauses and exponential backoff with jitter, with a tenacity example, though no Retry-After header (15). No SLA found (0). The voyage-4 series, voyage-context-4, voyage-code-4 and rerank-3 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.

## 2. Security & auth, 45 out of 100, up to 9.6 more on the total

Why it scored 45: Plain API keys from the dashboard, revocable, no scopes (20). The same key reaches the Batch and Files APIs, which can delete files, with no way to limit it (10 of 20). Returns vectors and scores, rerank returns the caller's own documents (10). No per-key audit log found (0 of 15). No security.txt, SOC 2 and HIPAA reports on a Vanta trust page (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.

## 3. 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-token prices public for every model (20). 200 million free tokens per current model with no card, though rate limits are very low until a payment method is added (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.

## 4. Schema & documentation, 61 out of 100, up to 6.3 more on the total

Why it scored 61: No public OpenAPI file found (0 of 25). llms.txt at docs.voyageai.com (10). The docs say which model fits general, code, finance, law, multimodal and chunk-in-context work, and when to set input_type (16 of 20). input_type, output_dimension and output_dtype take stated values, and per-request token caps are given per model (14 of 15). Examples in Python, TypeScript and curl, and an error-code page with a fix per status (13 of 15). The docs changelog has a single undated entry on rerank-lite-1, and model releases are dated only on the blog, newest 30 September 2026 (8 of 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.

## 5. Transparency & trust, 51 out of 100, up to 4.3 more on the total

Made of editorial 25, provenance 76.

Why it scored 51: Closed service under published terms, SDK MIT, voyage-4-nano weights on Hugging Face (15). Voyage keeps a perpetual licence to train on customer content unless an organisation admin opts out, the opt-out needs a payment method and can't be reversed in the dashboard, and content sent before the opt-out stays covered. The terms and FAQ agree with each other, but the default is the weakest in this batch (10 of 30). No deprecation policy or dated retirement notices found (0 of 20). The terms name Voyage AI Innovations, Inc. while the site says MongoDB, Inc., and no subprocessor list or data location found (0 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: voyageai.com, registered 2020-12-29 (5 years) (11 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 6. Maintenance & community, 78 out of 100, up to 1.9 more on the total

Why it scored 78: rerank-3 and rerank-3-lite went GA on 30 September 2026, two days ago (30). Three dated releases in the last 90 days, the Python SDK v0.5.0 on 10 July, voyage-code-4 on 13 August and rerank-3 on 30 September (20). voyageai-python had 9 open issues when we looked on 1 October, several from 2024 and 2025 with no visible maintainer reply, including a September 2025 report that contextualized_embed can return NaN arrays (5 of 25). Official Python and TypeScript SDKs (15). GitHub Actions for CI, lint and security scans, MIT, still version 0.x, last Python release 84 days ago (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.

## 7. Agent ergonomics, 98 out of 100, up to 0.3 more on the total

Why it scored 98: output_dimension 256 to 2048 and float, int8, uint8, binary or ubinary output per request, plus top_k on rerank (25). Up to 1,000 texts a call with token caps per model, and truncation on by default (20). The error-code page gives each status from 400 to 504 a meaning and a fix (18 of 20). Stateless calls, and the docs give exponential backoff with jitter (20). model and input are the only required fields, official Python and TypeScript SDKs (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.

## 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: whether status.voyageai.com is a working status page. It timed out on four attempts over three days, and no status page is linked from the footer or the docs index
- unchecked: the Vanta trust page's subprocessor list and hosting locations, which robots.txt keeps our reader out of
- The terms, which are the contract, name Voyage AI Innovations, Inc. MongoDB, Inc. appears only in the copyright line, and no page says whether MongoDB's DPA applies
- Whether rerank-2.5 and older models will be retired. The rerank-3 post keeps rerank-2.5 for existing users with no date

## Weaknesses

- Training on customer data is the default, and the opt-out needs a card on file and is one way
- No security.txt, and no status page linked or reachable
- Rate-limit tiers only begin once a payment method is added
- No public OpenAPI file, and releases are dated only on the blog
- Python SDK issues from 2024 and 2025 sit without a maintainer reply

## 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.

- Opt the organisation out of training before sending anything private. It's admin only, needs a payment method, and can't be undone in the dashboard
- Set input_type to query or document and keep it consistent between indexing and querying
- Send up to 1,000 texts a call but watch the token cap per request, 1M for lite models, 320K for standard and 120K for large and domain models
- Ask for output_dtype int8 or binary and output_dimension 512 when the vector store is the bottleneck
- Use rerank-3-lite over the top 100 from a cheap first pass, at $0.02 per million tokens

## What the review panel asked for

- Card-free training opt-out
- Publish an OpenAPI file
- Date the changelog entries

## 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.