Up to date
XiaomiMiMo MIT out 34 days ago
1.02T parameters (8 of 384 experts per token), text in, text out, 1M tokens of context. 5 providers serve it from $0.39 per million input tokens.
Verified 2026-10-25 Rank #273 in LLMs · score 84.8 method v1 Sources 4 Downloads 30d 59.3k · #368 of 619 tracked
Changed this week: 1 new quantized build on the Hub (9 in all).2026-10-24
Picked by rule from 17 tracked pieces: the lab's own post, the most shared, a field report, the latest.
MiMo-V2.6 models exhibited tool-call repetition where agents issued identical calls repeatedly, wasting context and stalling tasks.
The 2 picks above, plus these 15. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
5 providers with a live price, from $0.39 to $0.52 per million input tokens; the median is $0.43. One dot per provider, cheapest in teal.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Widely served (5 API providers); 9 quantized builds for local use.
Top Spaces running it: FineEnvs/multi-harness-rl ♥ 272 · FineEnvs/multi-harness-rl-slides ♥ 9 · quantid/measurement-lab ♥ 1
Weights only, an estimate: leave room for the KV cache. Method
About 615 GB at INT4, 1.23 TB at FP8, 2.46 TB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-09-21), amber triangles are 9 dated articles (each links to it), grey ticks are repo commits.
13 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported Agents’ Last Exam | 31.6 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported AutomationBench v1.0.6 | 53.1 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported CyberGym | 94 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported DeepSWE · Write code with agents | 71.9 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported ExploitGym | 17.8 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported GDPval-AA 2.1 | 1673 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported JobBench | 62 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported MiMo Code Bench | 63.2 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported MiMo Cyber Bench | 80.2 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported MiMo VisualCoding | 72.3 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported OSWorld-Verified | 82 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported ProgramBench | 26.5 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported SEC Bench Pro | 66.3 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported Terminal-Bench · Write code with agents | 34.9 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported Terminal-Bench · Write code with agents | 89.9 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported Terminal-Bench · Write code with agents | 47.9 | MiMo-V2.6 Pro | model author | 2026-10-07 |
| author-reported Toolathlon · Call tools and APIs | 76.9 | MiMo-V2.6 Pro | model author | 2026-10-07 |
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | MiMoV2ForCausalLM |
| Model type | mimo_v2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 10,647,286,656 · F32: 26,496 · F8_E4M3: 13,378,781,184 · U8: 1,000,190,509,056 |
| Size band | XXL |
| MoE | yes |
| Experts | 384 |
| Experts per token | 8 |
| Context length | 1,048,576 (config.json max_position_embeddings) |
| Languages | en, zh |
| License | MIT (mit) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · MiMoV2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · MiMoV2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for MiMo-V2.6-Pro-RL: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 2.46 TB · Cluster (≥640 GB) |
|---|---|
| FP8 (1 byte/param) | 1.23 TB · Cluster (≥640 GB) |
| INT4 (0.5 bytes/param) | 615 GB · Multi-GPU node (<640 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.
quantized: 9finetune: 3adapter: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| ggml-org/MiMo-V2.6-Pro-RL-GGUF | quantized | GGUF | — | — | mit | ggml-org | 7 | 5.6k |
| mlx-community/MiMo-V2.6-Pro-RL-mxfp4-q8 | quantized | mxfp4 | 1.02T | XXL | mit | mlx-community | 4 | 1.8k |
| inferencerlabs/MiMo-V2.6-Pro-RL-MLX-Q9 | quantized | MLX | — | — | mit | inferencerlabs | 1 | 1.2k |
| jarrelscy/MiMo-V2.6-Pro-RL-ARVQ-hybrid | quantized | — | 119B | XXL | mit | jarrelscy | 3 | 1k |
| DevQuasar/XiaomiMiMo.MiMo-V2.6-Pro-RL-GGUF | quantized | GGUF | — | — | — | DevQuasar | 1 | 1.1k |
| kernelpool/MiMo-V2.6-Pro-RL-MXFP4-GGUF | quantized | MXFP4 | — | — | mit | kernelpool | 1 | 331 |
| Aidev2006/Inquisitive-V2.6-Pro-RL | quantized | — | 1.02T | XXL | mit | Aidev2006 | 0 | 280 |
| Aidev2006/Inquisitive-V1-Pro-RL | quantized | — | 1.02T | XXL | mit | Aidev2006 | 0 | 259 |
| jarrelscy/MiMo-V2.6-Pro-EXL3 | finetune | EXL3 | — | — | — | jarrelscy | 3 | 0 |
| smalinin/MiMo-V2.6-Pro-RL-GGUF | quantized | GGUF | — | — | mit | smalinin | 0 | 119 |
| gghfexp/MiMo-V2.6-Pro-RL-control-vectors-v1 | adapter | GGUF | — | — | cc-by-nc-4.0 | gghfexp | 0 | 117 |
| immanuelpeter/MiMoViT-2.6-Pro | finetune | — | 739M | S | mit | immanuelpeter | 1 | 14 |
| jarrelscy/MiMo-V2.6-Pro-EXL3-FP4 | finetune | EXL3 | — | — | — | jarrelscy | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-10-09 | Align preprocessor config to training values (min/max_pixels, image_mean/std) | b61f218 |
| 2026-09-22 | Fix typo in report | 73875d0 |
| 2026-09-22 | Add ModelScope links | 438e9ce |
| 2026-09-21 | Add assets/architecture.png | 54b1049 |
| 2026-09-21 | Add model card | c165b4c |
| 2026-09-21 | Add files using upload-large-folder tool | a949623 |
| 2026-09-21 | Upload MiMo_V2_6_technical_report.pdf | 7b08b5a |
| 2026-09-21 | Add files using upload-large-folder tool | ff3db61 |
| 2026-09-21 | Add files using upload-large-folder tool | 6ab99ec |
| 2026-09-21 | Add files using upload-large-folder tool | d1838a9 |
| 2026-09-21 | initial commit | c25d97f |
Hosted endpoints observed for MiMo-V2.6-Pro-RL, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.39/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| DeepInfra openrouter | xiaomi/mimo-v2.6-pro | 1M | fp8 | $0.39 | $0.78 | — | — | yes | live | 2026-10-24 |
| Novita hf, openrouter | xiaomimimo/mimo-v2.6-pro | 1M | fp8 | $0.43 | $0.87 | 29.1 | 1614 ms | yes | live | 2026-10-25 |
| GMICloud openrouter | xiaomi/mimo-v2.6-pro | 1.0M | bf16 | $0.43 | $0.87 | — | — | yes | live | 2026-10-24 |
| Xiaomi openrouter | xiaomi/mimo-v2.6-pro | 1M | fp8 | $0.43 | $0.87 | — | — | yes | live | 2026-10-24 |
| StreamLake openrouter | xiaomi/mimo-v2.6-pro | 977K | unknown | $0.52 | $1.04 | — | — | yes | live | 2026-10-24 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:36 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:04 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
