{
  "fixes": {
    "slug": "nace-drex",
    "name": "Drex 1.5",
    "listing": "https://www.anchorterminal.com/tools/nace-drex",
    "markdown": "# Fix list: Drex 1.5\n\nFrom Anchor Terminal's listing at https://www.anchorterminal.com/tools/nace-drex, the October 2026 research run, assessed 10 October 2026. Grade B, 69.7 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 Drex 1.5: 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, 40 out of 100, up to 7.5 more on the total\n\nWhy it scored 40: No machine payment protocol on the API, and no x402, MPP or L402 in the pricing reference, the API reference or the OpenAPI document (0). Per-token prices are on the pricing reference without a login (20). A free tier needs no card, through $25 of signup credit after an emailed-code sign-in and a $5 monthly credit (20). Onboarding needs a browser sign-in and a key copied from the dashboard, so an agent cannot start alone (0). Total 40.\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. Security \u0026 auth, 65 out of 100, up to 6.1 more on the total\n\nWhy it scored 65: Bearer keys in the nace_sk_ format are revocable and can be rotated without downtime, and an account is capped at three active keys. No scopes are documented (20). No read-only mode or per-key limit was found, so least privilege stops at the account (10 of 20). The decision call returns only probabilities over options the caller supplies, with no free text. The checklist gives 10 to a tool that returns no untrusted content, and that is the award here (10). Each response carries a request id, and the dashboard Usage page exports a CSV by key and bucket (15). security.txt is valid until 9 October 2027. The SOC 2 Type 2 listing and the pentest report sit behind an access request and were not read. No bug bounty or public advisory page was found (10 of 20). Total 65.\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## 3. Reliability, 80 out of 100, up to 4 more on the total\n\nWhy it scored 80: A public status page with component history, a BetterStack page, names an API and dashboard component that serves /v1/systemone, and it shows 100 per cent uptime over 90 days (20). The history page returned the homepage, so no incident list was read. Drex has been public for 12 days, and the record is short for the model itself. The incident score is therefore 20 of 30. Clef, Celeris and OpenAI got 5 or 8 for records only a day or two old. Here the 90-day window covers Drex's 12 days with no incident shown, so the score is higher (20). Rate limits are published with numbers for the free and paid tiers (15). A 429 carries Retry-After and retry-after-ms, a 529 says to retry shortly, the docs advise capped backoff with jitter, and a decision call changes no stored state (15). No SLA for any tier was found, and the Trust Centre gives recovery objectives on request only (0). No beta or preview label was found on the API or its docs (10). Total 80.\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## 4. Transparency \u0026 trust, 63 out of 100, up to 3.2 more on the total\n\nMade of editorial 50, provenance 75.\n\nWhy it scored 63: The hosted service has a named licensor and terms with a stated forum, and the weights carry a published licence text. The licence is not open source, and it restricts commercial use above $1 million in revenue or funding and competing products (15 of 30). The data statements disagree. The developer page says no training on your data, the Terms bar training on identifiable inputs without consent, and the privacy notice says input and output may be used to improve the Services. No retention period for API inputs was found (10 of 30). The dated alias history and the retired-model rule give notice of changes, but no minimum notice period is published (15 of 20). The Trust Centre links a subprocessor list that was not read, and it names AWS and Azure without any regional commitment (10 of 20). Total 50.\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: nace.ai, no registry record we could read (0 of 15)\n- Changelog: not found (0 of 10)\n\n## 5. Maintenance \u0026 community, 65 out of 100, up to 3.1 more on the total\n\nWhy it scored 65: The last model change was on 28 September 2026, 12 days before this check, which earns the full 30. The docs list dated model entries in the last 90 days, drex-v1.0 on 24 September, the drex-latest moves on 27 and 28 September, and drex-v1.5 on 28 September (20). Responsiveness scores 0 because GitHub issues could not be read from this session (0 of 25, unchecked). The Python and TypeScript SDKs and the MCP server are current, each at 0.1.0 and published on 8 October 2026 (15). The repository has tests but no CI workflow in its tree, so CI scores 0 (0 of 10). Total 65.\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## 6. Agent ergonomics, 85 out of 100, up to 2.4 more on the total\n\nWhy it scored 85: A decision answer is one probability per option, and no tokens are generated, so the response is compact (25). Pagination does not apply. Limits on question count, body size and state tokens are stated, but there is no output-size control (15 of 20). Errors carry a type, a message and a request id, and some messages name the fix, such as the valid model ids (20). Retry guidance is documented with Retry-After, but /v1/systemone documents no idempotency key, and each successful call is debited once per request_id, so a retry may be billed again (10 of 20). The model is optional and defaults to drex-v1.5. Official SDKs exist in Python and TypeScript, and there is a CLI (15). Total 85.\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## 7. Schema \u0026 documentation, 88 out of 100, up to 2 more on the total\n\nWhy it scored 88: A public OpenAPI 3.1 document for the Drex API at console.nace.ai/openapi.json, with POST /v1/systemone and its request schema (25). llms.txt is served at drex.nace.ai/llms.txt. The docs have a Copy Markdown control, but the Markdown URLs we tried returned errors (10). The docs say what Drex is for and what it cannot do, since it generates no text and picks only from the options a caller supplies. No page lists languages or weak areas, and none says when not to use it (15 of 20). Questions are typed as noul, choice or score, with named options and a 512-question limit. The state is free-form JSON by design (13 of 15). Examples in curl, TypeScript, Python and the CLI, and an error reference with types, messages and headers (15). Pinned model ids and a dated alias history earn 10 of the 15 for versioning and changelog. No public changelog was found, so 5 is withheld (10). Total 88.\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## Deductions\n\nEach comes off the total. A fixed and documented problem counts for less at the next check.\n\n- 10 October 2026, read against the public board generated 28 September 2026. The site at www.nace.ai/drex states a Decision Index 0.2.1 score of 58.28 and says Drex 1.5 leads all 71 entries under 10B parameters. The 0.2.1 board has 70 entries, none of them Drex, and its top score is Surogate Rune 26B-A4B v3 at 57.44 (https://huggingface.co/spaces/multimodalart/jev-decision-index/resolve/main/data/index-v0.2.1.json). The Hugging Face card does say the Drex figure is Nace's own run, but the site does not. Deducted 2 as a misleading claim.\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: GitHub issue and pull request replies. The repository is not attached to this session, so its issues could not be read.\n- unchecked: the SOC 2 Type 2 report and the pentest report behind the Trust Centre access request.\n- unchecked: the subprocessor list and any data residency statement, behind the Trust Centre's SafeBase pages.\n- unchecked: the incident history. status.nace.ai/history returned the homepage, and only the 90-day uptime figure was read.\n- unchecked: console.nace.ai/llms.txt timed out. drex.nace.ai/llms.txt loaded.\n- The vendor says Drex 1.5 is the highest-scoring model under 10B parameters. Jev's size is unpublished and Drex is absent from both public boards, so the ranking cannot be checked.\n- The 58.28 on the 0.2.1 page equals Surogate Rune 26B-A4B v3's public-index score on the 0.3 board. The source of the figure is not known.\n- No announcement was found for the drex-latest moves on 27 and 28 September 2026, which the docs say are announced in advance.\n- Unsloth compatibility is not supported by any Nace page or repository read. The Hugging Face path unsloth/drex-v1.5 did not return a public page.\n- No Drex-specific service terms were found. The Terms are dated 5 May 2026, before the model was released.\n\n## Weaknesses\n\n- The rank 1 claim comes from Nace's own run on Decision Index 0.3.1, and Drex is not on the public board for 0.3 or 0.2.1.\n- The weights are under Open RAIL-M, which restricts commercial use above $1 million in revenue or funding and excludes competing products.\n- The data statements conflict. The developer page says no training on your data, while the privacy notice allows input and output to improve the Services.\n- No SLA, changelog or readable incident list for the model. Drex has been public for 12 days.\n- Sign-up needs an emailed code and a browser step. The $25 signup credit is granted once per mailbox and expires after three months.\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- Call https://console.nace.ai/v1/systemone with Authorization: Bearer nace_sk_..., and pin model drex-v1.5 rather than drex-latest in production.\n- Stop on 402 insufficient_credit or 402 payment_required until credit is topped up or the unpaid invoice is paid.\n- Retry 429 and 529 after retry-after-ms. On a 422 for an unknown model, use one of the ids the error message lists.\n- Keep state below 131,072 tokens and the body below 1,048,576 bytes, and send at most 512 questions per request.\n- Keep the key on a server. A key in browser code is exposed, and the /v1 routes send no Access-Control-Allow-Origin header.\n- Check usage.input_tokens and the usage CSV after retries, since /v1/systemone documents no idempotency key.\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": 69.7,
    "assessed": "2026-10-10",
    "run": "October 2026 research run",
    "categories": [
      {
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "score": 40,
        "maxGain": 7.5,
        "reason": "No machine payment protocol on the API, and no x402, MPP or L402 in the pricing reference, the API reference or the OpenAPI document (0). Per-token prices are on the pricing reference without a login (20). A free tier needs no card, through $25 of signup credit after an emailed-code sign-in and a $5 monthly credit (20). Onboarding needs a browser sign-in and a key copied from the dashboard, so an agent cannot start alone (0). Total 40.",
        "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": "security",
        "name": "Security \u0026 auth",
        "score": 65,
        "maxGain": 6.1,
        "reason": "Bearer keys in the nace_sk_ format are revocable and can be rotated without downtime, and an account is capped at three active keys. No scopes are documented (20). No read-only mode or per-key limit was found, so least privilege stops at the account (10 of 20). The decision call returns only probabilities over options the caller supplies, with no free text. The checklist gives 10 to a tool that returns no untrusted content, and that is the award here (10). Each response carries a request id, and the dashboard Usage page exports a CSV by key and bucket (15). security.txt is valid until 9 October 2027. The SOC 2 Type 2 listing and the pentest report sit behind an access request and were not read. No bug bounty or public advisory page was found (10 of 20). Total 65.",
        "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": 80,
        "maxGain": 4,
        "reason": "A public status page with component history, a BetterStack page, names an API and dashboard component that serves /v1/systemone, and it shows 100 per cent uptime over 90 days (20). The history page returned the homepage, so no incident list was read. Drex has been public for 12 days, and the record is short for the model itself. The incident score is therefore 20 of 30. Clef, Celeris and OpenAI got 5 or 8 for records only a day or two old. Here the 90-day window covers Drex's 12 days with no incident shown, so the score is higher (20). Rate limits are published with numbers for the free and paid tiers (15). A 429 carries Retry-After and retry-after-ms, a 529 says to retry shortly, the docs advise capped backoff with jitter, and a decision call changes no stored state (15). No SLA for any tier was found, and the Trust Centre gives recovery objectives on request only (0). No beta or preview label was found on the API or its docs (10). Total 80.",
        "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": "transparency",
        "name": "Transparency \u0026 trust",
        "score": 63,
        "maxGain": 3.2,
        "reason": "The hosted service has a named licensor and terms with a stated forum, and the weights carry a published licence text. The licence is not open source, and it restricts commercial use above $1 million in revenue or funding and competing products (15 of 30). The data statements disagree. The developer page says no training on your data, the Terms bar training on identifiable inputs without consent, and the privacy notice says input and output may be used to improve the Services. No retention period for API inputs was found (10 of 30). The dated alias history and the retired-model rule give notice of changes, but no minimum notice period is published (15 of 20). The Trust Centre links a subprocessor list that was not read, and it names AWS and Azure without any regional commitment (10 of 20). Total 50.",
        "blend": "editorial 50, provenance 75",
        "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": 65,
        "maxGain": 3.1,
        "reason": "The last model change was on 28 September 2026, 12 days before this check, which earns the full 30. The docs list dated model entries in the last 90 days, drex-v1.0 on 24 September, the drex-latest moves on 27 and 28 September, and drex-v1.5 on 28 September (20). Responsiveness scores 0 because GitHub issues could not be read from this session (0 of 25, unchecked). The Python and TypeScript SDKs and the MCP server are current, each at 0.1.0 and published on 8 October 2026 (15). The repository has tests but no CI workflow in its tree, so CI scores 0 (0 of 10). Total 65.",
        "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": "ergonomics",
        "name": "Agent ergonomics",
        "score": 85,
        "maxGain": 2.4,
        "reason": "A decision answer is one probability per option, and no tokens are generated, so the response is compact (25). Pagination does not apply. Limits on question count, body size and state tokens are stated, but there is no output-size control (15 of 20). Errors carry a type, a message and a request id, and some messages name the fix, such as the valid model ids (20). Retry guidance is documented with Retry-After, but /v1/systemone documents no idempotency key, and each successful call is debited once per request_id, so a retry may be billed again (10 of 20). The model is optional and defaults to drex-v1.5. Official SDKs exist in Python and TypeScript, and there is a CLI (15). Total 85.",
        "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": "schema",
        "name": "Schema \u0026 documentation",
        "score": 88,
        "maxGain": 2,
        "reason": "A public OpenAPI 3.1 document for the Drex API at console.nace.ai/openapi.json, with POST /v1/systemone and its request schema (25). llms.txt is served at drex.nace.ai/llms.txt. The docs have a Copy Markdown control, but the Markdown URLs we tried returned errors (10). The docs say what Drex is for and what it cannot do, since it generates no text and picks only from the options a caller supplies. No page lists languages or weak areas, and none says when not to use it (15 of 20). Questions are typed as noul, choice or score, with named options and a 512-question limit. The state is free-form JSON by design (13 of 15). Examples in curl, TypeScript, Python and the CLI, and an error reference with types, messages and headers (15). Pinned model ids and a dated alias history earn 10 of the 15 for versioning and changelog. No public changelog was found, so 5 is withheld (10). Total 88.",
        "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": "Domain age",
        "value": "nace.ai, no registry record we could read",
        "points": 0,
        "max": 15
      },
      {
        "label": "Changelog",
        "value": "not found",
        "points": 0,
        "max": 10
      }
    ],
    "deductions": [
      "10 October 2026, read against the public board generated 28 September 2026. The site at www.nace.ai/drex states a Decision Index 0.2.1 score of 58.28 and says Drex 1.5 leads all 71 entries under 10B parameters. The 0.2.1 board has 70 entries, none of them Drex, and its top score is Surogate Rune 26B-A4B v3 at 57.44 (https://huggingface.co/spaces/multimodalart/jev-decision-index/resolve/main/data/index-v0.2.1.json). The Hugging Face card does say the Drex figure is Nace's own run, but the site does not. Deducted 2 as a misleading claim."
    ],
    "unchecked": [
      "unchecked: GitHub issue and pull request replies. The repository is not attached to this session, so its issues could not be read.",
      "unchecked: the SOC 2 Type 2 report and the pentest report behind the Trust Centre access request.",
      "unchecked: the subprocessor list and any data residency statement, behind the Trust Centre's SafeBase pages.",
      "unchecked: the incident history. status.nace.ai/history returned the homepage, and only the 90-day uptime figure was read.",
      "unchecked: console.nace.ai/llms.txt timed out. drex.nace.ai/llms.txt loaded.",
      "The vendor says Drex 1.5 is the highest-scoring model under 10B parameters. Jev's size is unpublished and Drex is absent from both public boards, so the ranking cannot be checked.",
      "The 58.28 on the 0.2.1 page equals Surogate Rune 26B-A4B v3's public-index score on the 0.3 board. The source of the figure is not known.",
      "No announcement was found for the drex-latest moves on 27 and 28 September 2026, which the docs say are announced in advance.",
      "Unsloth compatibility is not supported by any Nace page or repository read. The Hugging Face path unsloth/drex-v1.5 did not return a public page.",
      "No Drex-specific service terms were found. The Terms are dated 5 May 2026, before the model was released."
    ],
    "weaknesses": [
      "The rank 1 claim comes from Nace's own run on Decision Index 0.3.1, and Drex is not on the public board for 0.3 or 0.2.1.",
      "The weights are under Open RAIL-M, which restricts commercial use above $1 million in revenue or funding and excludes competing products.",
      "The data statements conflict. The developer page says no training on your data, while the privacy notice allows input and output to improve the Services.",
      "No SLA, changelog or readable incident list for the model. Drex has been public for 12 days.",
      "Sign-up needs an emailed code and a browser step. The $25 signup credit is granted once per mailbox and expires after three months."
    ],
    "agentNotes": [
      "Call https://console.nace.ai/v1/systemone with Authorization: Bearer nace_sk_..., and pin model drex-v1.5 rather than drex-latest in production.",
      "Stop on 402 insufficient_credit or 402 payment_required until credit is topped up or the unpaid invoice is paid.",
      "Retry 429 and 529 after retry-after-ms. On a 422 for an unknown model, use one of the ids the error message lists.",
      "Keep state below 131,072 tokens and the body below 1,048,576 bytes, and send at most 512 questions per request.",
      "Keep the key on a server. A key in browser code is exposed, and the /v1 routes send no Access-Control-Allow-Origin header.",
      "Check usage.input_tokens and the usage CSV after retries, since /v1/systemone documents no idempotency key."
    ],
    "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"
  }
}