# Fix list: Jina Embeddings and Reranker From Anchor Terminal's listing at https://www.anchorterminal.com/tools/jina-embeddings, the October 2026 research run, assessed 1 October 2026. Grade C, 61.3 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 Jina Embeddings and Reranker: 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, 35 out of 100, up to 11.4 more on the total Why it scored 35: One Bearer key per account shared across Reader, Search, Embeddings, Reranker and the MCP server, revocable, no scopes (20). No destructive operations on the embedding API, but a leaked key drains a balance shared with Reader and Search (10 of 20). The embedding and rerank endpoints return vectors and scores, while the same MCP server also exposes web-reading tools that return untrusted pages, filterable by tag (5 of 15). No per-key log or audit trail found (0 of 15). No security.txt, the privacy statement is Elastic's, and no bug bounty or certification that names Jina found (0 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, 30 out of 100, up to 8.8 more on the total Why it scored 30: No x402, MPP or L402 (0). Tokens are prepaid in packs and the embeddings and reranker pages show no price in any currency, saying only that you may be charged in USD, EUR or other currencies, so no per-unit price without a login (0). The embeddings page says new users get an auto-generated key with free tokens, no card (20). That auto-generated key needs no signup form by the page's wording, but it comes from loading the website, not from an API, and we couldn't confirm whether a login sits in front of it (10 of 20). 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, 65 out of 100, up to 7 more on the total Why it scored 65: status.jina.ai on Atlassian Statuspage, with a component per embedding and reranker model (20). The history page renders only in a browser, but its RSS feed lists 26 automated incidents from 15 September to 1 October 2026, among them service outages on jina-code-embeddings-1500m on 18 and 21 September, a partial outage of jina-clip-v2 on 15 September and degradations of jina-embeddings-v4 and multi-embeddings, each marked resolved. Several majors on secondary models, and the feed stops at 15 September (5 of 30). Rate limits published per key tier, free 100 requests and 100,000 tokens a minute, paid 500 and 2 million, premium 5,000 and 50 million, though the LLM guide at docs.jina.ai gives 1 million and 5 million tokens for paid and premium (15). The embeddings page says a request over the limit 'returns HTTP 429 and should be retried with exponential backoff', and the OpenAPI document defines a TooManyRequests response. The API only reads, so retries are safe (15). No SLA found (0). The v5 embedding models and jina-reranker-v3.5 are on the production API, though v5-omni and reranker v3.5 have no status component (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, 61 out of 100, up to 3.4 more on the total Made of editorial 45, provenance 76. Why it scored 61: API closed, MCP server Apache-2.0, current model weights CC BY-NC 4.0 with a paid commercial licence (Jina On-Prem, sold by Elastic since 10 August 2026), not OSI (20 of 30). The terms, updated 4 May 2026, say Jina doesn't train on customer inputs and stores them only as needed to serve the request, and Elastic's data processing terms apply, though the privacy statement is Elastic's generic one (25 of 30). No deprecation policy or dated retirement notices found, and jina-embeddings-v4 is kept on the free tier only (0 of 20). German entity under Berlin courts, and no subprocessor list or data-location statement found for the API (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: jina.ai, registered 2020-01-20 (6 years) (11 of 15) - Changelog: not found (0 of 10) - security.txt: not found (0 of 10) ## 5. Maintenance & community, 62 out of 100, up to 3.3 more on the total Why it scored 62: The OpenAPI document is versioned 2026.09.17.0130 and the MCP server reached 1.10.0 on 18 September 2026, 14 days ago (30). Seven MCP version bumps from 1.5.0 on 11 August to 1.10.0 on 18 September, plus jina-reranker-v3.5 on 20 July (20). Issue replies not read, since GitHub's issue pages are closed to our reader (5 of 25). No official SDK package, the MCP server is the maintained client (5 of 15). The MCP repository has no CI workflow and no tests, only lint and type-check scripts, no tags or GitHub releases, and one committer (2 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. ## 6. Schema & documentation, 84 out of 100, up to 2.6 more on the total Why it scored 84: OpenAPI 3.1 document at api.jina.ai/openapi.json, version 2026.09.17.0130, covering embeddings, rerank, batch embeddings, classify and models (25). llms.txt at jina.ai/models/llms.txt and a Markdown guide written for models at docs.jina.ai (10). The model pages and llms.txt's selection principles say which model fits text, multimodal or code work and how many tokens an image costs (15 of 20). The spec types model, task and embedding_type as enums and bounds dimensions, and the MCP tools have typed JSON Schema inputs (14 of 15). curl examples on every model page, and the spec defines 400, 401, 403, 404, 409, 422, 429, 500, 503 and 504 responses for embeddings and rerank, though we couldn't read the full 429 body (12 of 15). Dated model releases and a dated spec version, but no API changelog (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. ## 7. Agent ergonomics, 86 out of 100, up to 2.3 more on the total Why it scored 86: dimensions truncation, float, binary, ubinary or base64 output, and top_n and return_documents on rerank (25). Batched inputs, an asynchronous batch embeddings endpoint, and auto-truncation inside the reranker's 131,072-token window (18 of 20). Error responses defined in the OpenAPI document for every 4xx and 5xx an agent would meet (16 of 20). Read-only calls, safe to retry, with exponential backoff documented for 429s and no idempotency key (18 of 20). No official SDK package, only HTTP and the MCP server (9 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. ## 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: incident history before 15 September 2026, since the history page renders only in a browser and the RSS feed stops there. Durations of the September incidents aren't in the feed - The price per token, which the public pages, llms.txt and docs.jina.ai don't show. - Whether the auto-generated key on jina.ai needs a login first. The page says new users get one and shows a separate login link - The embeddings page and docs.jina.ai give different token limits for the paid and premium tiers (2 million against 1 million, 50 million against 5 million a minute) - unchecked: issue replies on the MCP repository, whose GitHub issue pages are closed to our reader ## Weaknesses - No price per token in any currency on the public pages - One prepaid balance shared with Reader and Search, so a scraping job can drain the embedding budget - 26 automated incidents on the status feed from 15 September to 1 October 2026, and no status component for v5-omni or reranker v3.5 - No security.txt, no SLA and no API changelog - The MCP server has no CI or tests, and current weights are CC BY-NC 4.0 ## 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 the whole candidate set to rerank in one call. The 131K window on v3.5 fits hundreds of chunks - On a 429, back off exponentially. Limits count per key when a key is sent, per IP otherwise - Use /v1/batch/embeddings for large corpora rather than a loop of synchronous calls - Add include_tags=rerank on the MCP URL to load only sort_by_relevance and deduplicate_strings - Count image tokens before a big multimodal job, about 363 an image on v5-omni ## What the review panel asked for - Publish per-token prices - Split balances per API - Reconcile the two rate-limit tables - Add an API changelog ## 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.