MiMo-V2.6 Distill Qwen 9B

# Fix list: MiMo-V2.6 Distill Qwen 9B

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/mimo-v2-6-distill-qwen-9b, the October 2026 research run, assessed 21 October 2026. Grade C, 61 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 MiMo-V2.6 Distill Qwen 9B: 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. Risk, 60 out of 100, up to 12.3 more on the total

Why it scored 60: File safety 25, safetensors only and no repository code runs. Provenance 15, the files are in Xiaomi's MiMo organisation, which mimo.xiaomi.com links to, with a checksum on every file. Safety evaluation 0, the technical report has no safety results. Agent guidance 0, nothing on refusals or safeguards, though a seventh of the tuning data is offensive security. Licence risk 15, MIT and Apache 2.0 with no conditions. Security programme 5 of 10, Xiaomi's company security centre takes reports, but nothing links it to MiMo and no advisories were found.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-risk):

What could go wrong when you run it, and what the publisher has done about it. A higher score means less risk.

- 0 to 25, file safety. 25 for weights in safetensors or GGUF only, 15 when pickle files (.bin, .pt, .pth) sit beside them, 0 for pickle only. 10 off when the documented way to load it or to build its prompts runs code from the repository (`trust_remote_code`, or the publisher's scripts in place of a template).
- 0 to 15, provenance. 10 when the files come from the publisher's own organisation, the one its site or docs link to, and 5 for a checksum on every file.
- 0 to 20, safety evaluation. 20 when results are published and someone other than the publisher ran them or can rerun them, 10 for the publisher's own results, 0 for none.
- 0 to 15, guidance for agent use. 15 for documented prompt-injection guidance or an instruction hierarchy the model was trained to follow, 10 for a statement of what its trained refusals cover and the safeguards to add around it, 5 for general cautions, 0 for none.
- 0 to 15, licence risk. 15 to start, 5 off for each of these, a use policy or field-of-use limits beyond the law, terms the licensor can change or revoke, a revenue or user threshold, and a licence that changed without a dated note or whose link is broken. Not below 0.
- 0 to 10, a security programme. 5 for a way to report a flaw (security.txt, a disclosure policy or a security contact), 5 for advisories handled in public.

When our reference-machine runs start, measured robustness (prompt injection with AgentDojo, over-refusal with XSTest) joins Risk.

## 2. Reach, 70 out of 100, up to 11.5 more on the total

Why it scored 70: Memory 25 of 30, Xiaomi's only build is BF16, 18.85 GB or 22.6 GB with a fifth added, which fits 24 GB. Runtimes 20, it keeps Qwen3.5-9B's architecture, which released vLLM, SGLang, Transformers and llama.cpp run, while Ollama's entry is the base's. Builds 0, Xiaomi publishes no quantised build. Terms 15, MIT on the card over an Apache 2.0 base, both with nothing added. Download 10, not gated. Hosted 0, Xiaomi's API and OpenRouter serve the bigger MiMo-V2.6 models, not this one.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reach):

Whether an agent can run it, and where.

- 0 to 30, memory. The smallest build from the publisher or a runtime's own library that keeps 4-bit precision or better, at its file size plus a fifth for the runtime and a working context. 30 for 16 GB or less, 25 for 24 GB, 20 for 32 GB, 15 for 64 GB, 10 for what fits our reference machine (the GPU share of 128 GB of unified memory), 5 for 256 GB, 0 above.
- 0 to 25, runtimes. 5 for each of llama.cpp, Ollama, vLLM, SGLang, LM Studio and Hugging Face Transformers that runs the architecture in a released version, from the runtime's own code, library or documentation. A publisher's fork or an unreleased branch doesn't count. 25 at most.
- 0 to 15, builds from the publisher. 15 for quantised builds at two precisions or more (FP8, 4-bit, GGUF), 10 for one, 0 when only other people's quantisations exist.
- 0 to 15, terms of use. 15 for an OSI-approved licence with nothing added, 10 for a custom licence that lets anyone use it commercially for free below a threshold or under a use policy, 5 for research or non-commercial use only, 0 for no licence or unclear terms.
- 0 to 10, download. 10 when the files download with no account and no terms to accept, 5 for a click-through that needs an account, 0 when a person approves each request.
- 5, hosted too. The publisher's own API serves it, or at least two inference providers do, so an agent without the hardware can still call it.

When our reference-machine runs start, speed on that machine (tokens per second and time to first token at the build that fits) joins Reach.

## 3. Documentation and upkeep, 40 out of 100, up to 9.2 more on the total

Why it scored 40: Model card 5 of 20, intended use and parameters only. Training data 5, a domain mix with token counts but no sources or cut-off. Last change 20, the README changed 29 days ago. Responsiveness 0, Xiaomi answered none of the 14 discussions. Version policy 5, a first release. File health 5, no LICENSE file, a config that lists an MTP layer the weights lack, and template and tool-call reports waiting without a reply.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-upkeep):

- 0 to 20, the model card. Intended use, limitations, architecture, parameters, context length, languages and knowledge cut-off. 20 for all seven, 3 off for each that's missing.
- 0 to 20, training data. 20 for the data published or listed with sources and sizes, 15 for a data card that names the sources or the mix, the cut-off and the filtering, 10 for a general description with a cut-off, 5 without one, 0 for none.
- 0 to 20, the last change to the files, the template or the reference code. 20 within 30 days, 15 within 90, 10 within 180, 5 within a year, 0 older.
- 0 to 20, responsiveness. The publisher's replies to the 40 most recent questions and reports in the repository's discussions or issues. 20 for most, 10 for some, 5 for a few, 0 for none.
- 0 to 10, version policy. 10 for earlier versions kept in their own repositories with a dated note when one is replaced, 5 for earlier versions kept without notes or a first release, 0 when files were replaced in place without a note.
- 0 to 10, file health. 10 when config, tokenizer, template and card agree and no report of a broken file waits without a reply, 5 when one does, 0 when the files contradict the card.

## 4. Agent interface, 70 out of 100, up to 4.6 more on the total

Why it scored 70: Chat template 20, a Jinja file in the repository. Tool-calling format 10 of 25, the format is only in the template and the card names no tool parser. Trained tool calling 20, several calls a turn and agentic fine-tuning. Structured output 0, not documented. Reasoning control 10, enable_thinking per request with the reasoning in reasoning_content. Context 10, 262,144 tokens from the base's native length, though the card doesn't say.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-interface):

What a model and a developer need to talk to it. Speech, image and other models with no chat interface read the tool-calling and structured-output lines for what the model takes and returns (voices, timestamps, streaming, typed outputs), and the note says how.

- 0 to 20, the chat template. 20 for a template in the repository (Jinja, in the tokenizer config or its own file), 10 when the prompt format is only given as code you run.
- 0 to 25, the tool-calling format. 25 when the publisher documents how tools are declared and how calls come back, with the parser to use in a named runtime, 10 when only the template or a runtime's own recipe says it.
- 0 to 20, tool calling the model was trained for. 20 with several calls in one turn, 10 for one call at a time.
- 0 to 15, structured output. 15 when JSON or JSON Schema output is documented for the model with how to ask for it, 5 when the card claims it without saying how.
- 10, reasoning control. Thinking that can be turned off or budgeted per request, with where the reasoning text comes back.
- 0 to 10, context. 10 for 128K tokens or more trained, 5 for 32K to 127K. Context reached only by scaling the position encoding counts one band down.

## 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 Xiaomi's security centre covers MiMo models, since its page is drawn by script and no MiMo page links it. Credited 5 as a company reporting route
- The card's MIT label sits on a fine-tune of an Apache 2.0 model with no LICENSE or NOTICE file. Apache 2.0 asks redistributors to pass on its text, so the founder may want a line in the listing
- The context is credited from config.json and the base's native 262,144 tokens; Xiaomi doesn't say at what length it fine-tuned

## Weaknesses

- Only BF16 from Xiaomi, about 23 GB to run
- No safety results or agent guidance, with 11.0B tokens of offensive-security tuning
- No reply from Xiaomi to any of its 14 discussions, including template and tool-call reports
- No LICENSE file, and the card has no limitations, context, languages or cut-off
- The tool format and parser aren't documented on the card

## 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.

- Serve it as the card does, `sglang serve --model-path XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B --reasoning-parser mimo`, and add `--tool-call-parser mimo` for parsed tool calls
- SGLang's mimo reasoning parser expects thinking to be asked for, so pass `chat_template_kwargs` with `enable_thinking` true, as the card does, and read `reasoning_content`
- Calls come back as function and parameter blocks inside tool_call, the same shape vLLM's and SGLang's qwen3_coder parsers read
- The generation config samples at temperature 0.6, top_k 20 and top_p 0.95
- Images and video go in through Qwen3.5's processor; the template's audio tokens have no audio encoder behind them
- The MTP head isn't in the files, so skip MTP speculative decoding even though config.json lists a layer
- It's an SFT checkpoint meant for further RL, and discussions report runaway generation on llama.cpp, so cap max_tokens

## What the review panel asked for

- add a LICENSE file
- answer the template reports
- a 4-bit build from Xiaomi
- name a tool parser on the card
- state what refusals cover
- link a security contact from the card

## 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.