Unsloth by Unsloth

# Fix list: Unsloth

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/unsloth, the October 2026 research run, assessed 1 October 2026. Grade D, 51.7 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 Unsloth: 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. Reliability, 43 out of 100, up to 11.4 more on the total

Why it scored 43: Scored on the local-package checklist. Installs from PyPI (`unsloth`) or a Docker image, with Python 3.12 and 3.13 and Windows, Linux, WSL and macOS stated in the README (20). GitHub Actions workflows exist, but we didn't confirm a passing test run on main (10). 792 open issues and 472 open pull requests against a project that ships several releases a week; we didn't sample how many are crashes (8). Calendar versions (2026.9.12) with release notes that don't call out breaking changes (5). No 1.0 or stability declaration found, and the Studio's GitHub tags carry a -beta suffix (0).

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.

## 2. Security & auth, 35 out of 100, up to 11.4 more on the total

Why it scored 35: Read as local software. No account; Studio asks for an admin password once it's exposed beyond loopback and issues API keys for its server (20). Studio binds to loopback by default and `--disable-tools` turns off server-side tools, which are on by default when exposed (10). Those tools can act on untrusted input and the only guidance is the README's 'be careful' (5). No audit log found (0). No security.txt at unsloth.ai and no disclosure policy or advisories found (0).

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, 53 out of 100, up to 7.6 more on the total

Why it scored 53: Read as a framework an agent drives with code. A full SFT run is a notebook of a few dozen lines, and Studio adds an OpenAI-compatible endpoint (15). You choose what to save, an adapter of about 100 MB, merged 16-bit or GGUF, which controls output size (10). Error messages weren't tested in this run; the docs have a troubleshooting page (10). Training resumes from checkpoints through the Trainer, which is the nearest thing to a safe retry (10). Sensible defaults in every notebook, but Python only (8).

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, 34 out of 100, up to 5.8 more on the total

Made of editorial 40, provenance 27.

Why it scored 34: Apache-2.0 for the core and AGPL-3.0 for Studio, both OSI licences, and the README says which parts are which (30). Training data stays on your machine, but unsloth.ai has no privacy page (404) and nothing says what Studio or the desktop app sends home (10). No deprecation policy or dated notices found (0). `get_statistics` is exported from unsloth/models/_utils.py; we couldn't read its body, and no telemetry disclosure or opt-out found in the docs (0).

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):

- Legal entity named: not found (0 of 20)
- Domain age: unsloth.ai, no registry record we could read (0 of 15)
- Status page: not found (0 of 10)
- security.txt: not found (0 of 10)

## 5. Schema & documentation, 66 out of 100, up to 5.5 more on the total

Why it scored 66: Read as a framework. No API reference or typed contract published beyond the Python signatures; the docs teach through guides and notebooks (10). llms.txt at unsloth.ai/docs/llms.txt (10). The fine-tuning guide says when to pick LoRA, QLoRA or full fine-tuning and which hyperparameters to start from (15). Parameters such as `max_seq_length`, `load_in_4bit` and LoRA targets are explained in guides rather than a typed reference (8). Over 100 notebooks by model family, and a troubleshooting section (13). Dated releases on GitHub with notes, no separate changelog (10).

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. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: A free, self-hosted package with nothing to buy on the site or in the docs, so 20 + 20 + 20 for pricing, free use and no sign-up. No payment protocol (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.

## 7. Maintenance & community, 82 out of 100, up to 1.6 more on the total

Why it scored 82: unsloth 2026.9.12 on PyPI on 2026-09-28 (30). Fifteen PyPI releases between 25 August and 28 September (20). 792 open issues and 472 open pull requests; we didn't sample reply times (12). The PyPI package and Docker images are current (15). Workflows exist but CI status on main wasn't confirmed (5).

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.

## 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.

- What `get_statistics` sends, and whether it can be turned off, wasn't established; the function body was beyond what we could read.
- We didn't confirm the CI status on main or sample how many open issues are crashes or regressions.
- Whether Unsloth sells an enterprise or hosted plan wasn't rechecked this run; last week's check found none on the site.

## Weaknesses

- Not a hosted service; you bring and pay for the GPU
- Studio is AGPL-3.0, and its server-side tools are on by default when exposed
- 792 open issues and 472 open pull requests
- No legal entity in the terms, no privacy page and no security.txt
- Calendar versions with no breaking-change notes and no deprecation policy

## 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.

- Install with `uv pip install unsloth --torch-backend=auto` on a CUDA machine; the desktop app is for people
- Start from the notebook for the model family in unslothai/notebooks; it sets LoRA targets and the chat template
- Save the LoRA adapter while iterating and merge to 16-bit or GGUF only when you ship
- If Studio must be reachable by other agents, pass --disable-tools and keep it on 127.0.0.1 behind a tunnel
- Pin the exact unsloth version; releases land several times a week and don't flag breaking changes

## What the review panel asked for

- breaking-change notes per release
- a deprecation policy
- Explain what get_statistics sends

## 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.