{
  "fixes": {
    "slug": "pplx-decider",
    "name": "pplx-decider",
    "listing": "https://www.anchorterminal.com/tools/pplx-decider",
    "markdown": "# Fix list: pplx-decider\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/pplx-decider, the October 2026 research run, assessed 9 October 2026. Grade B, 62.9 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 pplx-decider: 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. Payments \u0026 pricing, 25 out of 100, up to 9.4 more on the total\n\nWhy it scored 25: No x402, MPP or L402 in the guide, the reference, the pricing page or llms.txt (0). The price is published without a login, $0.02 per million input tokens with free output and no per-request fee (20). Usage is prepaid, the billing guide asks for a credit card, and no free allowance was found (0). A person signs up in the console. The Stripe Projects route lets a coding agent provision a project, a key and credits from a terminal, but it needs a person's Stripe account with a payment method and a $10 minimum purchase, and the key-generation endpoint needs an existing key (5 of 20). The weights are free, and the hosted price is what this line grades.\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## 2. Reliability, 61 out of 100, up to 7.8 more on the total\n\nWhy it scored 61: Read as a hosted model API, the surface an agent would call. status.perplexity.com has an API component with uptime from July to October 2026 and an incident feed. It has no Decisions component (20). The Decisions API was released in October 2026, so its own record is days long. The feed names one API incident, on 13 August 2026, before the release and with no duration shown, so 8 of 30 for a clean record too short to count, the reading given to the other decision APIs in their first days. The guide states 10 requests a second per organisation on every plan, and a token limit with no number (13 of 15). 429 carries `Retry-After`, successful responses carry `x-ratelimit-*` headers, the guide says to retry 5xx with backoff, and a decision has no side effects (15). The API FAQ says Perplexity does not guarantee uptime or recovery time (0). The guide carries no beta label, but the description file is named `openapi-gateway-preview.json` at version 0.1.0 and the official Python SDK has no method for the endpoint (5 of 10).\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. Security \u0026 auth, 61 out of 100, up to 6.8 more on the total\n\nWhy it scored 61: Read for a model API. Project API keys are shown once, can be revoked, and can be generated and revoked through two documented endpoints with a rotation guide. No per-key scopes or spending limits were found (22 of 30). The API FAQ says query data sent through the API is not retained and not used for training. The Privacy and Security page states zero retention for the Chat Completions API only and does not name Decisions, and the governing terms were not read (12 of 20). The endpoint judges text and images that often come from users. It returns only probabilities and never fetches a URL, and the cookbook index describes code applying cut-offs before an action runs. No guidance on adversarial input was found in the guide or reference (5 of 15). Responses carry `x-request-id` and token usage, and admins see usage and per-key invoice lines in the console. No audit log was found (10 of 15). The docs name a SOC 2 Type II report and a trust centre, which is script-drawn and was not read. The SDK repository has a `SECURITY.md` with a private advisory form. No security.txt was found on the docs host and no bounty was found (12 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## 4. Transparency \u0026 trust, 52 out of 100, up to 4.2 more on the total\n\nMade of editorial 54, provenance 49.\n\nWhy it scored 52: The v1.1 weights are public under Apache-2.0 with a `NOTICE` naming the base model and revision, and the repository carries the training code, config, a data manifest with 349 sources and the vendor's evaluation file. The v1 repository states MIT for its `source/` folder. The v1.1 `source/` folder has no licence file of its own and its pyproject carries a private classifier. The hosted service is closed (24 of 30). The API FAQ says prompts are not retained or trained on. The Privacy and Security page limits that statement to the Chat Completions API, and the terms, privacy notice and any DPA were not read because www.perplexity.ai answered 403 (14 of 30). Retirements are announced in the changelog, with no dated policy or notice period (8 of 20). The FAQ says compute is on AWS in North America. No subprocessor list was read, since the trust centre is script-drawn (8 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- Legal entity named: not found (0 of 20)\n- Domain age: perplexity.ai, no registry record we could read (0 of 15)\n- Terms of service: published, but our reader couldn't read it (7 of 10)\n- Privacy policy: published, but our reader couldn't read it (7 of 10)\n- security.txt: could not be fetched (0 of 10)\n\n## 5. Agent ergonomics, 76 out of 100, up to 3.9 more on the total\n\nWhy it scored 76: Read as an API an agent calls for a decision. Answers are a few numbers per question, input may run to just under 262,144 tokens, and images travel inside `state`. No batch route or caching for the endpoint was found (20 of 25). The caller sets the whole output shape, with up to 128 questions, 255 options and 10 levels in a call (18 of 20). The error table gives a cause and an action for each status. 404 and 405 have empty bodies, 504 can be an HTML page, and an oversized image waits about a minute before a 504 where a 400 would be expected (14 of 20). A decision has no side effects, 429 names its wait, and the guide says identical requests can differ in the second decimal place (17 of 20). Three required fields and any HTTP client. The official Python SDK at 0.43.8, cloned on 9 October, has no Decisions method, and the TypeScript SDK was not read (7 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. Maintenance \u0026 community, 62 out of 100, up to 3.3 more on the total\n\nWhy it scored 62: Read for a hosted model with open weights. The v1.1 weights were committed on 5 October 2026 and the changelog carries the switch under October 2026 (30). The FAQ says retirements are announced in the changelog with a replacement. No minimum notice period was found. v1 still answers after v1.1 replaced it (5 of 12). Two versions in five days, with v1 kept callable and its price cut to match, which is too short a history to judge churn (5 of 8). The changelog has several entries a month, and support runs through a community forum, a GitHub discussion repository and two email addresses. Replies were not sampled (10 of 15). The official Python SDK is current, with 0.43.8 on 7 October 2026 and three releases in 30 days, but has no Decisions method (7 of 15). The weights repository pins its requirements and ships a `uv.lock`. Its server code reads a `playground.html` that the v1.1 repository does not include, and the card still calls the repository private (5 of 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## 7. Schema \u0026 documentation, 90 out of 100, up to 1.6 more on the total\n\nWhy it scored 90: An OpenAPI 3.0.3 description of `POST /decisions` is embedded in the reference page and listed in llms.txt as `openapi-gateway-preview.json`. We read the embedded copy (25). llms.txt indexes the guide, the reference and four cookbook pages, each with a Markdown twin (10). The guide says what a decision model is, where to use it in place of a chat model and what it does not do. The model card adds serving caveats. No page lists languages or known weak areas (16 of 20). Questions are a union discriminated on `type`, with 1 to 128 questions, 1 to 255 options, 1 to 10 levels and unknown fields rejected. `state` and `instructions` accept any string, object or array by design (13 of 15). Examples in Python, TypeScript and curl with a real response, and a status table for 400, 401, 404, 405, 413, 429 and 5xx. The guide says `error.code` changes type between errors (14 of 15). Versioned model IDs that must be named on every request, a `/v1` path and a changelog with both Decisions entries, dated by month only (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## 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: www.perplexity.ai answered 403 with a bot check, so the terms of service, the privacy notice, any API-specific terms or DPA and the legal entity were not read. `provenance.terms` and `provenance.privacy` are left out. Read them before any probe is run\n- unchecked: trust.perplexity.ai is script-drawn, so the SOC 2 Type II report, the subprocessor list and any security contact were not read\n- unchecked: the vendor announcement in the lead was not found on a page we could read. The price, the licence and the leaderboard result were each confirmed from other sources. The words tying the index to Hugging Face were not\n- The lead called the benchmark the Hugging Face Decision Index 0.3. It is the Space multimodalart/jev-decision-index, labelled Decision Index 0.3.1, under a personal account whose profile lists membership of the `huggingface` organisation. The Space calls itself unofficial and community-maintained\n- The leaderboard result is the index's measurement on one NVIDIA RTX PRO 6000, and 80 per cent of the Full score comes from private tests nobody outside can inspect. We have not run the model\n- The model's entry on the index carries the note that repository access was by request, and the card still says private, while the repository is public and not gated today. When it opened was not established\n- The training manifest lists 349 sources, most from tasksource. We compared their names with the index's public benchmarks and found no shared name, which is not a contamination audit. The licences of those sources were not reviewed\n- unchecked: the TypeScript SDK, the Hugging Face discussion threads, reply times on the community forum, and whether any host other than Perplexity serves the model\n- The guide carries no beta label, but the description file is named preview at version 0.1.0. Whether Perplexity treats the endpoint as generally available was not stated\n- The hosted API accepts just under 262,144 input tokens. The published inference code rejects a question over 8,192 tokens. How the hosted service differs from the published code was not established\n\n## Weaknesses\n\n- The Decisions API was released in October 2026, so its incident record is days long, and Perplexity's FAQ says it gives no uptime guarantee\n- The official Python SDK at 0.43.8 has no Decisions method. The guide uses a plain HTTP client\n- An image over 2,048 tiles is not rejected. The request waits about a minute and returns 504\n- The zero-retention statement on the Privacy and Security page names only the Chat Completions API\n- The bundled server in the weights repository caps a question at 8,192 tokens and reads a `playground.html` the v1.1 repository does not include\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- Send `Authorization: Bearer`. A key in `x-api-key` returns 401 on this endpoint.\n- Set `model` on every request. A missing or unknown model returns 400, and the guide shows a `-latest` name rejected.\n- Resize images to 2,048 tiles of 32 by 32 pixels or fewer before sending. Larger images end in a 504 after about a minute.\n- Check the HTTP status before parsing an error. 404 and 405 have empty bodies, 504 can be HTML, and `error.code` changes type.\n- For the weights, use the repository's `autojev` code. The card says default causal inference does not reproduce the evaluated setup.\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": 62.9,
    "assessed": "2026-10-09",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 25,
        "maxGain": 9.4,
        "reason": "No x402, MPP or L402 in the guide, the reference, the pricing page or llms.txt (0). The price is published without a login, $0.02 per million input tokens with free output and no per-request fee (20). Usage is prepaid, the billing guide asks for a credit card, and no free allowance was found (0). A person signs up in the console. The Stripe Projects route lets a coding agent provision a project, a key and credits from a terminal, but it needs a person's Stripe account with a payment method and a $10 minimum purchase, and the key-generation endpoint needs an existing key (5 of 20). The weights are free, and the hosted price is what this line grades.",
        "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": "reliability",
        "name": "Reliability",
        "score": 61,
        "maxGain": 7.8,
        "reason": "Read as a hosted model API, the surface an agent would call. status.perplexity.com has an API component with uptime from July to October 2026 and an incident feed. It has no Decisions component (20). The Decisions API was released in October 2026, so its own record is days long. The feed names one API incident, on 13 August 2026, before the release and with no duration shown, so 8 of 30 for a clean record too short to count, the reading given to the other decision APIs in their first days. The guide states 10 requests a second per organisation on every plan, and a token limit with no number (13 of 15). 429 carries `Retry-After`, successful responses carry `x-ratelimit-*` headers, the guide says to retry 5xx with backoff, and a decision has no side effects (15). The API FAQ says Perplexity does not guarantee uptime or recovery time (0). The guide carries no beta label, but the description file is named `openapi-gateway-preview.json` at version 0.1.0 and the official Python SDK has no method for the endpoint (5 of 10).",
        "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": "security",
        "name": "Security \u0026 auth",
        "score": 61,
        "maxGain": 6.8,
        "reason": "Read for a model API. Project API keys are shown once, can be revoked, and can be generated and revoked through two documented endpoints with a rotation guide. No per-key scopes or spending limits were found (22 of 30). The API FAQ says query data sent through the API is not retained and not used for training. The Privacy and Security page states zero retention for the Chat Completions API only and does not name Decisions, and the governing terms were not read (12 of 20). The endpoint judges text and images that often come from users. It returns only probabilities and never fetches a URL, and the cookbook index describes code applying cut-offs before an action runs. No guidance on adversarial input was found in the guide or reference (5 of 15). Responses carry `x-request-id` and token usage, and admins see usage and per-key invoice lines in the console. No audit log was found (10 of 15). The docs name a SOC 2 Type II report and a trust centre, which is script-drawn and was not read. The SDK repository has a `SECURITY.md` with a private advisory form. No security.txt was found on the docs host and no bounty was found (12 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": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 52,
        "maxGain": 4.2,
        "reason": "The v1.1 weights are public under Apache-2.0 with a `NOTICE` naming the base model and revision, and the repository carries the training code, config, a data manifest with 349 sources and the vendor's evaluation file. The v1 repository states MIT for its `source/` folder. The v1.1 `source/` folder has no licence file of its own and its pyproject carries a private classifier. The hosted service is closed (24 of 30). The API FAQ says prompts are not retained or trained on. The Privacy and Security page limits that statement to the Chat Completions API, and the terms, privacy notice and any DPA were not read because www.perplexity.ai answered 403 (14 of 30). Retirements are announced in the changelog, with no dated policy or notice period (8 of 20). The FAQ says compute is on AWS in North America. No subprocessor list was read, since the trust centre is script-drawn (8 of 20).",
        "blend": "editorial 54, provenance 49",
        "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": "ergonomics",
        "name": "Agent ergonomics",
        "score": 76,
        "maxGain": 3.9,
        "reason": "Read as an API an agent calls for a decision. Answers are a few numbers per question, input may run to just under 262,144 tokens, and images travel inside `state`. No batch route or caching for the endpoint was found (20 of 25). The caller sets the whole output shape, with up to 128 questions, 255 options and 10 levels in a call (18 of 20). The error table gives a cause and an action for each status. 404 and 405 have empty bodies, 504 can be an HTML page, and an oversized image waits about a minute before a 504 where a 400 would be expected (14 of 20). A decision has no side effects, 429 names its wait, and the guide says identical requests can differ in the second decimal place (17 of 20). Three required fields and any HTTP client. The official Python SDK at 0.43.8, cloned on 9 October, has no Decisions method, and the TypeScript SDK was not read (7 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": "maintenance",
        "name": "Maintenance \u0026 community",
        "score": 62,
        "maxGain": 3.3,
        "reason": "Read for a hosted model with open weights. The v1.1 weights were committed on 5 October 2026 and the changelog carries the switch under October 2026 (30). The FAQ says retirements are announced in the changelog with a replacement. No minimum notice period was found. v1 still answers after v1.1 replaced it (5 of 12). Two versions in five days, with v1 kept callable and its price cut to match, which is too short a history to judge churn (5 of 8). The changelog has several entries a month, and support runs through a community forum, a GitHub discussion repository and two email addresses. Replies were not sampled (10 of 15). The official Python SDK is current, with 0.43.8 on 7 October 2026 and three releases in 30 days, but has no Decisions method (7 of 15). The weights repository pins its requirements and ships a `uv.lock`. Its server code reads a `playground.html` that the v1.1 repository does not include, and the card still calls the repository private (5 of 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"
      },
      {
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "score": 90,
        "maxGain": 1.6,
        "reason": "An OpenAPI 3.0.3 description of `POST /decisions` is embedded in the reference page and listed in llms.txt as `openapi-gateway-preview.json`. We read the embedded copy (25). llms.txt indexes the guide, the reference and four cookbook pages, each with a Markdown twin (10). The guide says what a decision model is, where to use it in place of a chat model and what it does not do. The model card adds serving caveats. No page lists languages or known weak areas (16 of 20). Questions are a union discriminated on `type`, with 1 to 128 questions, 1 to 255 options, 1 to 10 levels and unknown fields rejected. `state` and `instructions` accept any string, object or array by design (13 of 15). Examples in Python, TypeScript and curl with a real response, and a status table for 400, 401, 404, 405, 413, 429 and 5xx. The guide says `error.code` changes type between errors (14 of 15). Versioned model IDs that must be named on every request, a `/v1` path and a changelog with both Decisions entries, dated by month only (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"
      }
    ],
    "provenance": [
      {
        "label": "Legal entity named",
        "value": "not found",
        "points": 0,
        "max": 20
      },
      {
        "label": "Domain age",
        "value": "perplexity.ai, no registry record we could read",
        "points": 0,
        "max": 15
      },
      {
        "label": "Terms of service",
        "value": "published, but our reader couldn't read it",
        "points": 7,
        "max": 10
      },
      {
        "label": "Privacy policy",
        "value": "published, but our reader couldn't read it",
        "points": 7,
        "max": 10
      },
      {
        "label": "security.txt",
        "value": "could not be fetched",
        "points": 0,
        "max": 10
      }
    ],
    "unchecked": [
      "unchecked: www.perplexity.ai answered 403 with a bot check, so the terms of service, the privacy notice, any API-specific terms or DPA and the legal entity were not read. `provenance.terms` and `provenance.privacy` are left out. Read them before any probe is run",
      "unchecked: trust.perplexity.ai is script-drawn, so the SOC 2 Type II report, the subprocessor list and any security contact were not read",
      "unchecked: the vendor announcement in the lead was not found on a page we could read. The price, the licence and the leaderboard result were each confirmed from other sources. The words tying the index to Hugging Face were not",
      "The lead called the benchmark the Hugging Face Decision Index 0.3. It is the Space multimodalart/jev-decision-index, labelled Decision Index 0.3.1, under a personal account whose profile lists membership of the `huggingface` organisation. The Space calls itself unofficial and community-maintained",
      "The leaderboard result is the index's measurement on one NVIDIA RTX PRO 6000, and 80 per cent of the Full score comes from private tests nobody outside can inspect. We have not run the model",
      "The model's entry on the index carries the note that repository access was by request, and the card still says private, while the repository is public and not gated today. When it opened was not established",
      "The training manifest lists 349 sources, most from tasksource. We compared their names with the index's public benchmarks and found no shared name, which is not a contamination audit. The licences of those sources were not reviewed",
      "unchecked: the TypeScript SDK, the Hugging Face discussion threads, reply times on the community forum, and whether any host other than Perplexity serves the model",
      "The guide carries no beta label, but the description file is named preview at version 0.1.0. Whether Perplexity treats the endpoint as generally available was not stated",
      "The hosted API accepts just under 262,144 input tokens. The published inference code rejects a question over 8,192 tokens. How the hosted service differs from the published code was not established"
    ],
    "weaknesses": [
      "The Decisions API was released in October 2026, so its incident record is days long, and Perplexity's FAQ says it gives no uptime guarantee",
      "The official Python SDK at 0.43.8 has no Decisions method. The guide uses a plain HTTP client",
      "An image over 2,048 tiles is not rejected. The request waits about a minute and returns 504",
      "The zero-retention statement on the Privacy and Security page names only the Chat Completions API",
      "The bundled server in the weights repository caps a question at 8,192 tokens and reads a `playground.html` the v1.1 repository does not include"
    ],
    "agentNotes": [
      "Send `Authorization: Bearer`. A key in `x-api-key` returns 401 on this endpoint.",
      "Set `model` on every request. A missing or unknown model returns 400, and the guide shows a `-latest` name rejected.",
      "Resize images to 2,048 tiles of 32 by 32 pixels or fewer before sending. Larger images end in a 504 after about a minute.",
      "Check the HTTP status before parsing an error. 404 and 405 have empty bodies, 504 can be HTML, and `error.code` changes type.",
      "For the weights, use the repository's `autojev` code. The card says default causal inference does not reproduce the evaluated setup."
    ],
    "recheck": "https://www.anchorterminal.com/builders/#disputes"
  },
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-10-10",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.4",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run"
  }
}