{
  "fixes": {
    "slug": "convai-laya",
    "name": "Laya",
    "listing": "https://www.anchorterminal.com/tools/convai-laya",
    "markdown": "# Fix list: Laya\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/convai-laya, the October 2026 research run, assessed 1 October 2026. Grade B, 69.2 out of 100.\n\nThis 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.\n\nFor a coding agent working on Laya: 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.\n\n## 1. Security \u0026 auth, 57 out of 100, up to 7.5 more on the total\n\nWhy it scored 57: Read as software you run. No account. `laya-serve` binds to 0.0.0.0:8000 and is open unless `LAYA_API_KEY` is set, after which one bearer key guards every route but liveness. The NixOS module reads the key through systemd credentials (10 of 30). No write actions. Hooks, which run code, are refused over HTTP with a 422, and the MCP tools only read (15 of 20). SECURITY.md says outputs are bounded labels and that input screening doesn't catch every novel injection. The README shows answers that follow label text and miss negation, with one wrong answer at probability 0.9998. That's a written warning without a mitigation (9 of 15). Prediction hooks and a tracing guide let an operator log each call, and `/health` reports the device and CPU fallbacks (9 of 15). SECURITY.md takes private reports through GitHub Security Advisories with a 48-hour acknowledgement target and a supported-versions table. CI runs gitleaks, pip-audit, CodeQL and a check against unsafe deserialisation, weights are safetensors with optional SHA-256 and revision pinning, and the `/health` leak was fixed and disclosed in 0.3.23. No bug bounty, and no security.txt on convaiinnovations.com (14 of 20).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-security):\n\n- 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.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 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.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.\n\nModels 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.\n\n## 2. Reliability, 65 out of 100, up to 7 more on the total\n\nWhy it scored 65: Scored on the local-package checklist, since Laya is a model you run. `pip install laya` from PyPI, at 0.3.23, with Python 3.10 to 3.13 stated (20). Public CI runs the Python suite on all four versions and the TypeScript SDK on Node 20 and 22, and the latest runs on main had passed when we looked on 2 October (25). The repository page showed 61 open issues and 81 open pull requests on 2 October, against about 1,150 commits from 146 authors since 18 September. Issues from 19 September were still open, among them #21 (the quickstart couldn't be reproduced) and #20 (Armenian text routed to the English checkpoint), and the README documents a negation failure (#377). We didn't check reply times (12 of 25). 26 tagged releases between 19 September and 1 October, with notes that call out behaviour changes such as `/health` hiding details from callers without the key and a new ONNX quantisation default, but no changelog file (8 of 15). Still 0.x and marked beta in the package metadata (0).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):\n\nHosted APIs, MCP servers, models and platforms.\n\n- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 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.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 10, the surface agents use is generally available, not beta or preview.\n\nLocal packages, SDKs, frameworks and stdio MCP servers.\n\n- 20, installs from an official package with supported runtimes stated.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 15, version 1.0 or later, or declared stable.\n\nProtocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.\n\n## 3. Payments \u0026 pricing, 60 out of 100, up to 5 more on the total\n\nWhy it scored 60: Free Apache-2.0 software with nothing to buy, so 20 + 20 + 20 for pricing, free use and no sign-up. No payment protocol (0). The maintainer takes donations through Buy Me A Coffee.\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):\n\nThe published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).\n\n- 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.\n- 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.\n- 20, a free tier or trial that doesn't need a card.\n- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).\n\nPayment 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.\n\nOpen-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.\n\n## 4. Schema \u0026 documentation, 80 out of 100, up to 3.3 more on the total\n\nWhy it scored 80: Read for a model you serve yourself. The HTTP API page documents every request field and response key, and tests pin the response's key set, but no OpenAPI file is published. The MCP server's 8 tools are typed Python functions (18 of 25). The docs site has no llms.txt (404), and the docs are Markdown in the repository (5 of 10). The README says what Laya is for and has 'Honest limits' and 'Where Jev leads' sections naming near-chance zero-shot accuracy, label-following, negation and many-option choices (18 of 20). Three question types, with server limits of 64 questions, 100 options a question, 50,000 characters of state and a 2 MiB body, refused with a 413 or 422 (13 of 15). Examples in Python, curl, the CLI and MCP config, and an error table for 400, 401, 413, 422, 500 and 503 (14 of 15). Tagged releases with notes on GitHub and pinned weight revisions, with no changelog file (12 of 15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):\n\nAPIs and MCP servers.\n\n- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 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.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 15, versioning and a public changelog.\n\nModels 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.\n\n## 5. Agent ergonomics, 80 out of 100, up to 3.3 more on the total\n\nWhy it scored 80: Read as an API an agent calls for a decision. Answers are small and `output_tokens` is always 0. Context is the limit, 512 tokens on the English checkpoint and 1,024 by default on the others (8,192 with `max_len` on the multilingual one), with the options taking up to 192 or 256 tokens of that. `predict_long` windows longer documents (15 of 25). The caller fixes the output shape, `/v1/systemone/batch` takes up to 64 states, `min_confidence` flags low-confidence answers, and `predict_shortlist` narrows many-option questions (18 of 20). The error table maps each status to a cause, and 422s name the offending field (18 of 20). Calls are stateless. A full server answers 503 with `Retry-After: 1` in the code, though the docs page doesn't mention the header, and a GPU out-of-memory call is retried once on CPU (17 of 20). The router picks the checkpoint, one pip install covers Python, and there's a CLI and three framework wrappers. The TypeScript SDK in the repository wasn't on npm (404) (12 of 15).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):\n\n- 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).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.\n\nModels 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.\n\n## 6. Transparency \u0026 trust, 62 out of 100, up to 3.3 more on the total\n\nMade of editorial 70, provenance 53.\n\nWhy it scored 62: Apache-2.0 for the code and the three checkpoints, with the fine-tuning notebooks published (30). SECURITY.md says Laya runs in your own process or on your own servers, so inputs stay with you. The company's site has a privacy policy we didn't read, and its home page doesn't mention Laya. Nothing says what the demo Space keeps (20 of 30). SECURITY.md's supported-versions table puts 0.3.x as active, 0.2.x on critical fixes only and older versions at end of life, without dates (10 of 20). We found no telemetry code in the package and no statement either way. Weights download through the Hugging Face Hub client (10 of 20).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):\n\n- 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.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).\n\nThe other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.\n\nProvenance checks not met in full (half of this category, computed from checked facts):\n\n- Domain age: convaiinnovations.com, no registry record we could read (0 of 15)\n- Status page: not found (0 of 10)\n- security.txt: not found (0 of 10)\n\n## 7. Maintenance \u0026 community, 83 out of 100, up to 1.5 more on the total\n\nWhy it scored 83: Read for an open-weight model. 0.3.23 on 1 October 2026 (30). 26 releases since 19 September (20). Pull requests are merged daily, 64 from 21 contributors in 0.3.23 alone, but 61 issues and 81 pull requests were open and we didn't check reply times (15 of 25). The Python package is current. The TypeScript SDK isn't on npm, and the MCP server isn't in the official MCP registry (8 of 15). CI passes on main, with Dependabot, actions pinned by commit and trusted publishing to PyPI (10).\n\nThe checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):\n\n- 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.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 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.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.\n\nModels are read for deprecation notice periods and model churn rather than release counts.\n\n## What we couldn't check\n\nWhat 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.\n\n- unchecked: issue and pull-request reply times\n- The repository page showed 61 open issues while the issues page listed 7 for our reader. We used the repository page's count\n- Whether `laya-ts` will reach npm. The README says releases come from `laya-ts-v*` tags, and none exists\n- Hugging Face reports 0 downloads and 4,168 likes for convaiinnovations/laya. The download count looks like a counting gap, so it isn't used\n- Cloudflare's scores for Laya don't say which checkpoint it ran\n- unchecked: convaiinnovations.com's terms, privacy policy and domain registration date\n\n## Weaknesses\n\n- Base checkpoints score 0.362 and 0.352 on the maintainers' typed-decisions benchmark against a 0.318 random baseline, so it needs fine-tuning\n- Choice options share a 192 or 256-token budget, and the README reports 0.425 on Banking77's 77 labels\n- 512 tokens of context on the English checkpoint and 1,024 by default on the others\n- `laya-serve` listens on 0.0.0.0 with no key unless `LAYA_API_KEY` is set\n- 26 releases in 13 days, still 0.x and marked beta, with 61 open issues and 81 open pull requests\n\n## What costs an agent a turn today\n\nThe notes we give agents before they call it. Each one is a workaround an agent shouldn't need.\n\n- Set `LAYA_API_KEY` before starting `laya-serve`. It listens on every interface by default\n- Gate on `answer_confidence`, not `confidence`, which measures entropy and doesn't match Jev's field\n- Shortlist choice questions with more than about 20 options using `predict_shortlist` or the `laya_shortlist` tool\n- Use semantic or opaque labels such as `A` and `B`, not `yes` and `no`, in choice questions. The checkpoints can follow the label text\n- Pass `model=\"multilingual\"` and `max_len=8192` for long documents. The English checkpoint stops at 512 tokens\n\n## What the review panel asked for\n\n- dates on support windows\n- a changelog file\n- cost per 1,000 questions table\n- schema token count for MCP tools\n\n## When it's done\n\nSend 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.\n",
    "grade": "B",
    "score": 69.2,
    "assessed": "2026-10-01",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "security",
        "name": "Security \u0026 auth",
        "score": 57,
        "maxGain": 7.5,
        "reason": "Read as software you run. No account. `laya-serve` binds to 0.0.0.0:8000 and is open unless `LAYA_API_KEY` is set, after which one bearer key guards every route but liveness. The NixOS module reads the key through systemd credentials (10 of 30). No write actions. Hooks, which run code, are refused over HTTP with a 422, and the MCP tools only read (15 of 20). SECURITY.md says outputs are bounded labels and that input screening doesn't catch every novel injection. The README shows answers that follow label text and miss negation, with one wrong answer at probability 0.9998. That's a written warning without a mitigation (9 of 15). Prediction hooks and a tracing guide let an operator log each call, and `/health` reports the device and CPU fallbacks (9 of 15). SECURITY.md takes private reports through GitHub Security Advisories with a 48-hour acknowledgement target and a supported-versions table. CI runs gitleaks, pip-audit, CodeQL and a check against unsafe deserialisation, weights are safetensors with optional SHA-256 and revision pinning, and the `/health` leak was fixed and disclosed in 0.3.23. No bug bounty, and no security.txt on convaiinnovations.com (14 of 20).",
        "checklist": [
          "- 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.\n- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.\n- 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.\n- 0 to 15, audit logs or per-call visibility for the operator.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-security"
      },
      {
        "key": "reliability",
        "name": "Reliability",
        "score": 65,
        "maxGain": 7,
        "reason": "Scored on the local-package checklist, since Laya is a model you run. `pip install laya` from PyPI, at 0.3.23, with Python 3.10 to 3.13 stated (20). Public CI runs the Python suite on all four versions and the TypeScript SDK on Node 20 and 22, and the latest runs on main had passed when we looked on 2 October (25). The repository page showed 61 open issues and 81 open pull requests on 2 October, against about 1,150 commits from 146 authors since 18 September. Issues from 19 September were still open, among them #21 (the quickstart couldn't be reproduced) and #20 (Armenian text routed to the English checkpoint), and the README documents a negation failure (#377). We didn't check reply times (12 of 25). 26 tagged releases between 19 September and 1 October, with notes that call out behaviour changes such as `/health` hiding details from callers without the key and a new ONNX quantisation default, but no changelog file (8 of 15). Still 0.x and marked beta in the package metadata (0).",
        "checklist": [
          "Hosted APIs, MCP servers, models and platforms.",
          "- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).\n- 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.\n- 15, rate limits documented with numbers.\n- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.\n- 10, an SLA published for any paid tier.\n- 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.\n- 25, a public CI and test suite, passing on the default branch.\n- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).\n- 15, semver discipline and breaking changes called out in a changelog.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-reliability"
      },
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 60,
        "maxGain": 5,
        "reason": "Free Apache-2.0 software with nothing to buy, so 20 + 20 + 20 for pricing, free use and no sign-up. No payment protocol (0). The maintainer takes donations through Buy Me A Coffee.",
        "checklist": [
          "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.\n- 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.\n- 20, a free tier or trial that doesn't need a card.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-payments"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 80,
        "maxGain": 3.3,
        "reason": "Read for a model you serve yourself. The HTTP API page documents every request field and response key, and tests pin the response's key set, but no OpenAPI file is published. The MCP server's 8 tools are typed Python functions (18 of 25). The docs site has no llms.txt (404), and the docs are Markdown in the repository (5 of 10). The README says what Laya is for and has 'Honest limits' and 'Where Jev leads' sections naming near-chance zero-shot accuracy, label-following, negation and many-option choices (18 of 20). Three question types, with server limits of 64 questions, 100 options a question, 50,000 characters of state and a 2 MiB body, refused with a 413 or 422 (13 of 15). Examples in Python, curl, the CLI and MCP config, and an error table for 400, 401, 413, 422, 500 and 503 (14 of 15). Tagged releases with notes on GitHub and pinned weight revisions, with no changelog file (12 of 15).",
        "checklist": [
          "APIs and MCP servers.",
          "- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).\n- 10, llms.txt or Markdown docs served for agents.\n- 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.\n- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.\n- 0 to 15, examples and documented error responses.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-schema"
      },
      {
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "score": 80,
        "maxGain": 3.3,
        "reason": "Read as an API an agent calls for a decision. Answers are small and `output_tokens` is always 0. Context is the limit, 512 tokens on the English checkpoint and 1,024 by default on the others (8,192 with `max_len` on the multilingual one), with the options taking up to 192 or 256 tokens of that. `predict_long` windows longer documents (15 of 25). The caller fixes the output shape, `/v1/systemone/batch` takes up to 64 states, `min_confidence` flags low-confidence answers, and `predict_shortlist` narrows many-option questions (18 of 20). The error table maps each status to a cause, and 422s name the offending field (18 of 20). Calls are stateless. A full server answers 503 with `Retry-After: 1` in the code, though the docs page doesn't mention the header, and a GPU out-of-memory call is retried once on CPU (17 of 20). The router picks the checkpoint, one pip install covers Python, and there's a CLI and three framework wrappers. The TypeScript SDK in the repository wasn't on npm (404) (12 of 15).",
        "checklist": [
          "- 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).\n- 20, pagination, filtering and output-size controls.\n- 20, actionable, documented error responses, codes and messages an agent can recover from.\n- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-ergonomics"
      },
      {
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 62,
        "maxGain": 3.3,
        "reason": "Apache-2.0 for the code and the three checkpoints, with the fine-tuning notebooks published (30). SECURITY.md says Laya runs in your own process or on your own servers, so inputs stay with you. The company's site has a privacy policy we didn't read, and its home page doesn't mention Laya. Nothing says what the demo Space keeps (20 of 30). SECURITY.md's supported-versions table puts 0.3.x as active, 0.2.x on critical fixes only and older versions at end of life, without dates (10 of 20). We found no telemetry code in the package and no statement either way. Weights download through the Hugging Face Hub client (10 of 20).",
        "blend": "editorial 70, provenance 53",
        "checklist": [
          "- 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.\n- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).\n- 0 to 20, a deprecation policy or notices with dates.\n- 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."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-transparency"
      },
      {
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 83,
        "maxGain": 1.5,
        "reason": "Read for an open-weight model. 0.3.23 on 1 October 2026 (30). 26 releases since 19 September (20). Pull requests are merged daily, 64 from 21 contributors in 0.3.23 alone, but 61 issues and 81 pull requests were open and we didn't check reply times (15 of 25). The Python package is current. The TypeScript SDK isn't on npm, and the MCP server isn't in the official MCP registry (8 of 15). CI passes on main, with Dependabot, actions pinned by commit and trusted publishing to PyPI (10).",
        "checklist": [
          "- 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.\n- 20, at least three releases or dated changelog entries in the last 90 days.\n- 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.\n- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).\n- 10, package health, current dependencies and CI.",
          "Models are read for deprecation notice periods and model churn rather than release counts."
        ],
        "checklistUrl": "https://www.anchorterminal.com/benchmark/#checklist-maintenance"
      }
    ],
    "provenance": [
      {
        "label": "Domain age",
        "value": "convaiinnovations.com, no registry record we could read",
        "points": 0,
        "max": 15
      },
      {
        "label": "Status page",
        "value": "not found",
        "points": 0,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "unchecked": [
      "unchecked: issue and pull-request reply times",
      "The repository page showed 61 open issues while the issues page listed 7 for our reader. We used the repository page's count",
      "Whether `laya-ts` will reach npm. The README says releases come from `laya-ts-v*` tags, and none exists",
      "Hugging Face reports 0 downloads and 4,168 likes for convaiinnovations/laya. The download count looks like a counting gap, so it isn't used",
      "Cloudflare's scores for Laya don't say which checkpoint it ran",
      "unchecked: convaiinnovations.com's terms, privacy policy and domain registration date"
    ],
    "weaknesses": [
      "Base checkpoints score 0.362 and 0.352 on the maintainers' typed-decisions benchmark against a 0.318 random baseline, so it needs fine-tuning",
      "Choice options share a 192 or 256-token budget, and the README reports 0.425 on Banking77's 77 labels",
      "512 tokens of context on the English checkpoint and 1,024 by default on the others",
      "`laya-serve` listens on 0.0.0.0 with no key unless `LAYA_API_KEY` is set",
      "26 releases in 13 days, still 0.x and marked beta, with 61 open issues and 81 open pull requests"
    ],
    "agentNotes": [
      "Set `LAYA_API_KEY` before starting `laya-serve`. It listens on every interface by default",
      "Gate on `answer_confidence`, not `confidence`, which measures entropy and doesn't match Jev's field",
      "Shortlist choice questions with more than about 20 options using `predict_shortlist` or the `laya_shortlist` tool",
      "Use semantic or opaque labels such as `A` and `B`, not `yes` and `no`, in choice questions. The checkpoints can follow the label text",
      "Pass `model=\"multilingual\"` and `max_len=8192` for long documents. The English checkpoint stops at 512 tokens"
    ],
    "requests": [
      {
        "text": "dates on support windows",
        "reviews": 1
      },
      {
        "text": "a changelog file",
        "reviews": 1
      },
      {
        "text": "cost per 1,000 questions table",
        "reviews": 1
      },
      {
        "text": "schema token count for MCP tools",
        "reviews": 1
      }
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-04",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.3",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}