MiMo-V2.6-Flash-RL vs Qwen3-4B-Instruct-2507

MiMo-V2.6-Flash-RL 59

Model card 14, Intended use, the architecture, the parameters, the context and the languages (in the metadata) are on the card. Training data 5, A description of the training method with no data sources, sizes or cut-off. Last change 20, the last commits, on 28 October 2026, updated the card, 13 days before this check. Responsiveness 5, Of the 13 most recent discussions that aren't pull requests (21 September to 6 November 2026), 3 have a reply from a XiaomiMiMo member, which is a few. Version policy 5, MiMo-V2.5 stays in its own repositories and V2.6 came out in new ones, with no dated note on the older cards that we read. File health 10, Config, template and card agree on the architecture and the 1M context, and two early faults (invalid JSON in the draft model's config, preprocessor values) were fixed in commits that say so. 14 + 5 + 20 + 5 + 5 + 10 = 59.

Qwen3-4B-Instruct-2507 26

Model card 11 of 20, intended use, architecture, parameters and context are there, but limitations, named languages and a knowledge cut-off aren't. Training data 5, the cited Qwen3 report describes the data in general terms without a cut-off. Last change 0, the files last changed on 17 September 2025 (Hub lastModified on 4 November), more than a year ago. Responsiveness 0, counted by script on 4 November, Qwen answered none of the 20 questions and reports. Version policy 5, the older Qwen3-4B stays in its own repository without a note. File health 5, the files agree with the card in substance but reports wait without a reply. 11 + 5 + 0 + 0 + 5 + 5 = 26.

MiMo-V2.6-Flash-RL 65

Memory 5, the smallest build at 4 bits or better is ggml-org's MXFP4 GGUF, 168.93 GB with its projector, and 168.93 GB × 1.2 = 202.7 GB, within 256 GB. Xiaomi's own files are 177.74 GB. ggml-org's Q2_K is below 4 bits and isn't read. Runtimes 20, Released vLLM 0.33.0 and SGLang 0.5.23 have the mimo_v2 model and the mimo parsers, Transformers 5.22.2 has a mimo_v2_flash model, and llama.cpp b12242 has the architecture, with a GGUF from its own organisation. Builds 10, one precision from Xiaomi, the 4-bit files it releases. Terms 15, MIT in the card's metadata, an OSI licence. Download 10, the Hub API reports gated false on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, the card says it is on Xiaomi's MiMo Open Platform API and on OpenRouter, and Novita is live for it on Hugging Face. 5 + 20 + 10 + 15 + 10 + 5 = 65.

Qwen3-4B-Instruct-2507 90

Memory 30, LM Studio's and Ollama's 4-bit builds are 2.3 to 2.5 GB, about 3 GB with a fifth added, inside 16 GB. Runtimes 25, released Transformers, vLLM, SGLang and llama.cpp run the qwen3 architecture, Ollama's released build pins a llama.cpp that has it, and LM Studio's catalogue carries it. Builds 10 of 15, Alibaba's only quantised build is FP8. Terms 15, Apache 2.0 with nothing added. Download 10, neither repository is gated. Hosted 0, on 4 November 2026 the Hub's provider mapping showed only Featherless AI with status error, Nscale was gone and Alibaba's Model Studio doesn't list it. 30 + 25 + 10 + 15 + 10 + 0 = 90.

MiMo-V2.6-Flash-RL 45

File safety 5, the weights are safetensors, but the dflash folder's mask_embedding.pt is a pickle beside them (15). Provenance 15, the files sit in the XiaomiMiMo organisation, which mimo.xiaomi.com links to per the Distill listing (10). Safety evaluation 0, no safety results on the card. Agent guidance 5, General cautions only. Licence risk 15, MIT has no use policy, threshold or change clause. Security programme 5, Xiaomi runs a company security centre for reporting flaws, as read for the Distill listing (5). 5 + 15 + 0 + 5 + 15 + 5 = 45.

Qwen3-4B-Instruct-2507 55

File safety 25, safetensors only and nothing from the repository runs to load it or build prompts. Provenance 15, the files are in the Qwen organisation that Qwen's GitHub links to, with a checksum on every file. Safety evaluation 0, none published. Agent guidance 0, nothing on prompt injection, refusals or safeguards. Licence risk 15, plain Apache 2.0 that hasn't changed. Security programme 0, no security.txt, disclosure policy or security contact for Qwen was found, and Alibaba's group security centre couldn't be checked, so it's scored as absent. Files unchanged on 4 November, and QwenLM/Qwen3 still has no SECURITY.md. 25 + 15 + 0 + 0 + 15 + 0 = 55.

MiMo-V2.6-Flash-RL 85

Chat template 20, chat_template.jinja in the repository (3,868 characters), with a copy in tokenizer_config.json. Tool-calling format 25, the card names the parser to use in two runtimes: the mimo tool-call parser in SGLang and in vLLM, with auto tool choice, and both ship it in a release. Trained tool calling 20, the card describes reinforcement learning on coding and general agent tasks across several harnesses, and the template renders every call in a turn. Structured output 0, Nothing on the card about JSON or schema-constrained output. Reasoning control 10, the template has an enable_thinking switch and keeps reasoning_content, and the card names the mimo reasoning parser for SGLang and vLLM, which returns the reasoning apart from the answer. Context 10, 1M tokens per the card's model summary, and 1,048,576 positions in config.json. 20 + 25 + 20 + 0 + 10 + 10 = 85.

Qwen3-4B-Instruct-2507 75

Chat template 20, a Jinja template in tokenizer_config.json. Tool-calling format 25, Qwen's own Function Calling guide documents the Hermes-style declaration and how calls come back, and names vLLM's hermes parser. Trained tool calling 20, the template allows one or more calls a turn. Structured output 0, not documented for the model, and Qwen's docs only point to vLLM's generic guided decoding. Reasoning control 0, this edition has no thinking. Context 10, 262,144 tokens native. Template and config unchanged on 4 November. 20 + 25 + 20 + 0 + 0 + 10 = 75.