# Fix list: SambaCloud From Anchor Terminal's listing at https://www.anchorterminal.com/tools/sambanova, the October 2026 research run, assessed 8 October 2026. Grade B, 66.4 out of 100. This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public. For a coding agent working on SambaCloud: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published. ## 1. Payments & pricing, 40 out of 100, up to 7.5 more on the total Why it scored 40: No machine payment protocol (0). Per-token prices on the pricing page and in `/v1/models`, both read without a login, such as gpt-oss-120b at $0.22 in and $0.59 out per 1M tokens (20). Free tier with no payment method (20). A person signs up in the console to get a key (0). The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments): The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/). - 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which. - 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login. - 20, a free tier or trial that doesn't need a card. - 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API). Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied. Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol. ## 2. Security & auth, 58 out of 100, up to 7.4 more on the total Why it scored 58: Model reading. Bearer keys from the console, up to 25 per user and shown once. No scopes, project limits or rotation guidance were found, and the key travels in a header, never in the URL (18 of 30). The terms let SambaNova access, store and process Customer Content solely to supply the service and for no other purposes, which rules out training without naming it (18 of 20). No retention period is published. The product page says SambaCloud never sees or collects prompts, while the terms grant a right to store content and to use query logs (7 of 15). The console has a Usage page. No request logs or audit log were found in the docs (5 of 15). The trust centre states SOC 2 Type 2 and ISO 27001. No security.txt, disclosure policy or bug bounty was found, and the SDK's SECURITY.md sends service reports to a general address (10 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-security): - 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option. - 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions. - 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10. - 0 to 15, audit logs or per-call visibility for the operator. - 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public. Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing. ## 3. Agent ergonomics, 67 out of 100, up to 5.4 more on the total Why it scored 67: Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Function calling with `tool_choice` to force a function, on six listed models, and a July 2026 fix for tool calls that were buffered when streaming (17 of 20). `json_object` and `json_schema` output, best effort only, with a known failure on DeepSeek-V3.2 recorded in July 2026 (8 of 15). Automatic prompt caching on the MiniMax models only, with cached input on MiniMax-M3 at a tenth of the input price, a 4,096-token minimum and no guaranteed persistence (10 of 15). 131,072-token context on four of six models, 1M on the MiniMax-M3 preview, and output capped at 7,168 tokens on DeepSeek and 3,072 on Llama 3.3 70B (8 of 15). No batch API found in the docs index or the OpenAPI document (0 of 10). Official SDKs for Python and TypeScript, and the OpenAI and Anthropic clients work with a changed base URL (10). Errors carry `type`, `param`, `code` and `request_id` (14 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics): - 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries). - 20, pagination, filtering and output-size controls. - 20, actionable, documented error responses, codes and messages an agent can recover from. - 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations. - 15, sensible defaults, few required parameters, and official SDKs in at least two languages. Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs. ## 4. Transparency & trust, 63 out of 100, up to 3.2 more on the total Made of editorial 44, provenance 81. Why it scored 63: Closed service with clear terms that name the entity and the governing law, and Apache-2.0 SDKs and OpenAPI document (15). The privacy policy is dated 27 May 2023 and does not name SambaCloud or API inputs, no DPA or retention period is published, and the product page's claim that prompts are never seen or collected sits beside terms that allow storing content and using query logs (10 of 30). A deprecations page with notice and removal dates back to April 2025, though it has no entry for Mistral Large 3 or MiniMax-M2.7 (14 of 20). The terms name Stripe as payment processor and the privacy policy names regions in general terms. No sub-processor list or data location for inference was found (5 of 20). The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency): - 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms. - 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors). - 0 to 20, a deprecation policy or notices with dates. - 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted). The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two. Provenance checks not met in full (half of this category, computed from checked facts): - Domain age: sambanova.ai, registered 2017-12-18 (8 years) (11 of 15) - Terms of service: read, states 4 of the 7 things a reader expects, and has 1 clause that costs points (5.4 of 10) - security.txt: not found (0 of 10) ## 5. Schema & documentation, 81 out of 100, up to 3.1 more on the total Why it scored 81: Public OpenAPI 3.1.1 document, version 1.2.0, with 10 operations and 169 schemas, last updated 17 September 2026 (25). `llms.txt` at docs.sambanova.ai/docs/llms.txt and a Markdown copy of each page (10). 717 descriptions in the document, with feature pages for function calling, Responses, Messages and caching. When not to use a route is rarely stated (15 of 20). Typed parameters with 152 enums, but JSON schema output is best effort and `strict: true` has no effect (10 of 15). Code examples in Python and TypeScript, and an errors page with 13 codes and a recovery suggestion for each (13 of 15). The spec is versioned and release notes are dated, but the newest SambaCloud entry is 7 July 2026 and the models page disagrees with `/v1/models` (8 of 15). The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema): APIs and MCP servers. - 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool). - 10, llms.txt or Markdown docs served for agents. - 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference. - 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs. - 0 to 15, examples and documented error responses. - 15, versioning and a public changelog. Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference. ## 6. Reliability, 85 out of 100, up to 3 more on the total Why it scored 85: Hosted reading. Status page at status.sambanova.ai on Statuspage, with components for the API gateway, the Playground and each model (20). The incident feed lists 25 incidents since launch, the latest an 8-minute Playground login fault on 25 June 2026, so nothing is posted for the last 90 days (30). Limits published per model and tier, 20 requests a minute and 20 a day on the free tier, 60 and 12,000 on the Developer tier (15). `x-ratelimit-*` headers on every response, the error table says to retry `queue_full` after a delay, and the SDKs retry 429 and server errors twice with backoff. No `Retry-After` header was found in the docs or the OpenAPI document (10 of 15). No SLA found on the legal page or in the terms (0). Three production models are generally available, three are previews (10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability): Hosted APIs, MCP servers, models and platforms. - 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own). - 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so. - 15, rate limits documented with numbers. - 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved. - 10, an SLA published for any paid tier. - 10, the surface agents use is generally available, not beta or preview. Local packages, SDKs, frameworks and stdio MCP servers. - 20, installs from an official package with supported runtimes stated. - 25, a public CI and test suite, passing on the default branch. - 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered). - 15, semver discipline and breaking changes called out in a changelog. - 15, version 1.0 or later, or declared stable. Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors. ## 7. Maintenance & community, 77 out of 100, up to 2 more on the total Why it scored 77: Model reading. The OpenAPI document and both SDKs were updated on 17 September 2026, and a MiniMax-M3 component joined the status page on 1 September (30). Production models get a notice of at least two to three weeks, and the recorded removals had 14 days or less (5 of 12). No removal is recorded in the last 90 days, but 13 model ids left between 9 March and 9 June 2026, and MiniMax-M2.7 is in the docs but not in the live catalogue (4 of 8). The SambaCloud release notes stop at 7 July 2026. The community forum has staff replies and a post dated 8 October 2026 (10 of 15). The Python SDK repository has no open issues and holds release pull requests only, so replies could not be judged (5 of 10). Python 1.13.0 and TypeScript 1.10.0, both 17 September 2026 (15). CI, release-please, a lock file and gitleaks scanning (8 of 10). The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance): - 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older. - 20, at least three releases or dated changelog entries in the last 90 days. - 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15. - 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models). - 10, package health, current dependencies and CI. Models are read for deprecation notice periods and model churn rather than release counts. ## Deductions Each comes off the total. A fixed and documented problem counts for less at the next check. - 2026-10-08. The models and rate-limit pages list `MiniMax-M2.7` as a production model, and the July 2026 release notes say `Mistral-Large-3-675B-Instruct-2512` was promoted to production. Neither is on the pricing page or in `/v1/models`, and the deprecations page records neither. Two points, since a keyed call was not made to confirm the ids fail (https://docs.sambanova.ai/docs/en/models/sambacloud-models, https://api.sambanova.ai/v1/models). ## What we couldn't check What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it. - unchecked: whether calls to `MiniMax-M2.7` and `Mistral-Large-3-675B-Instruct-2512` fail. No keyed call was made. The finding rests on the pricing page and the keyless `/v1/models` response - unchecked: the trust centre beyond its page description (it renders with JavaScript), so the SOC 2 and ISO 27001 reports, any sub-processor list and any disclosure policy were not read - unchecked: the console after login, so key revocation, usage views and any request logs are not confirmed - No SLA, DPA, retention period or sub-processor list was found on the legal page. Whether they exist on request is not established - The speed claim behind `inference.fast` is the vendor's, citing Artificial Analysis. We did not measure it - The privacy policy predates SambaCloud's September 2024 launch and does not name it. It is linked as the only privacy policy the vendor publishes - The terms bar publishing benchmark results without written consent, which matters before our probes run ## Weaknesses - Production models get a notice of two to three weeks, and 13 model ids were removed between 9 March and 9 June 2026 - The docs list `MiniMax-M2.7` as a production model, while the pricing page and `/v1/models` did not list it on 8 October 2026 - `strict: true` on a JSON schema is accepted and has no effect, and the July 2026 notes record failing structured output on DeepSeek-V3.2 - Keys carry no scopes or project limits, and no key rotation guidance was found in the reviewed documentation - The privacy policy is dated 27 May 2023, and no DPA, retention period or sub-processor list was found on the legal page ## What costs an agent a turn today The notes we give agents before they call it. Each one is a workaround an agent shouldn't need. - Call `GET https://api.sambanova.ai/v1/models` at start-up and use only ids it returns. The docs name models the endpoint no longer lists - Read `x-ratelimit-remaining-requests` and `x-ratelimit-remaining-requests-day` on every response. The free tier allows 20 requests a day per model - Treat 429 `queue_full` and 503 `maintenance` as retryable after a delay, and 410 `model_deprecated` as a signal to change model - Validate JSON output yourself. Schema enforcement is best effort and `strict: true` changes nothing - Check `max_completion_tokens` per model. DeepSeek-V3.1 caps output at 7,168 tokens and Llama 3.3 70B at 3,072 ## When it's done Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.