Up to date
XiaomiMiMo MIT out 6 months ago
1.02T parameters (8 of 384 experts per token), text in, text out, 1M tokens of context. 4 providers serve it from $0.43 per million input tokens.
Verified 2026-10-25 Rank #386 in LLMs · score 79.9 method v1 Sources 4 Downloads 30d 12.9k · #487 of 619 tracked
Changed this week: 1 provider no longer listed: Novita.2026-10-23
Picked by rule from 17 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Xiaomi released MiMo-V2.5-Pro-UltraSpeed, a 1-trillion-parameter model achieving 1000 tokens/s decode speed via collaboration with TileRT.
Xiaomi MiMo permanently renovates its pricing system effective May 27, 2026, with API prices reduced by up to 99% and token usage increased to 5-8 times the original.
Xiaomi released MiMo-V2.5-Pro, an open-source coding model that finished a university compiler project in 4.3 hours and built a video editor using 1,868 tool calls.
Xiaomi engineers optimized the MiMo-V2.5 series for inference by implementing a tiered caching system that supports Hybrid Sliding Window Attention.
The 5 picks above, plus these 12. 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.
3 providers with a live price, from $0.43 to $1.00 per million input tokens; the median is $0.48. 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.
Served by 4 API providers; 9 quantized builds for local use.
Top Spaces running it: quickgrid/Tokenizer-Visualizer ♥ 6 · SrRooT/Tokenizer-Visualizer ♥ 1 · quantid/measurement-lab ♥ 1
Weights only, an estimate: leave room for the KV cache. Method
About 614 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 (2026-04-27), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
10 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.
MiMo-V2.5-Pro is an open-source Mixture-of-Experts language model with 1.02T total parameters.[1] The model is designed for agentic, complex software engineering, and long-horizon tasks.[2] It features a hybrid attention architecture and supports a context length of up to 1M tokens.[3]
Generated 2026-10-07 15:05 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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: 9,450,733,440 · F32: 162,817,920 · F8_E4M3: 1,013,569,290,240 |
| 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-V2.5-Pro: 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) | 614 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: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| unsloth/MiMo-V2.5-Pro-GGUF | quantized | GGUF | — | — | mit | unsloth | 20 | 20.8k |
| bartowski/MiMo-V2.5-Pro-GGUF | quantized | GGUF | — | — | mit | bartowski | 2 | 8.6k |
| AesSedai/MiMo-V2.5-Pro-GGUF | quantized | GGUF | — | — | — | AesSedai | 12 | 5.9k |
| kernelpool/MiMo-V2.5-Pro-6bit | quantized | — | 1.02T | XXL | mit | kernelpool | 1 | 2.9k |
| ji-farthing/MiMo-V2.5-Pro-DFlash-draft-ik-llama-GGUF | quantized | GGUF | — | — | mit | ji-farthing | 6 | 2.6k |
| DevQuasar/XiaomiMiMo.MiMo-V2.5-Pro-GGUF | quantized | GGUF | — | — | — | DevQuasar | 0 | 2.7k |
| gghfez/MiMo-V2.5-Pro-ik_llama-unfused-GGUF | quantized | GGUF | — | — | — | gghfez | 2 | 1.5k |
| fanwu103/MiMo-V2.5-Pro-MXFP4v2 | quantized | — | 554B | XXL | mit | fanwu103 | 0 | 105 |
| ckoh04/MiMo-V2.5-Pro-FP8-active-slice-8experts | quantized | -FP8 | 43.7B | XL | mit | ckoh04 | 0 | 77 |
| Qapdex/MiMo-V2.5-QEdge | finetune | — | — | — | mit | Qapdex | 1 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-07-09 | Update README.md | 21d1ecf |
| 2026-05-08 | update wechat qr code and discord invite link (#12) | f5953f1 |
| 2026-04-28 | Add ModelScope url | a75207d |
| 2026-04-27 | Upload 14 files | 7208274 |
| 2026-04-27 | Add files using upload-large-folder tool | 0df734e |
| 2026-04-27 | Add files using upload-large-folder tool | 258de6b |
| 2026-04-27 | initial commit | d5590c0 |
Hosted endpoints observed for MiMo-V2.5-Pro, 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: $0.30/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| GMICloud openrouter | xiaomi/mimo-v2.5-pro | 1.0M | bf16 | $0.30 | $0.61 | — | — | — | status -5 | 2026-10-24 |
| AtlasCloud openrouter | xiaomi/mimo-v2.5-pro | 1000K | fp8 | $0.43 | $0.87 | — | — | yes | live | 2026-10-24 |
| DigitalOcean openrouter | xiaomi/mimo-v2.5-pro | 256K | unknown | $0.48 | $1.80 | — | — | yes | live | 2026-10-24 |
| DeepInfra hf | XiaomiMiMo/MiMo-V2.5-Pro | 1M | — | $1.00 | $3.00 | 2.1 | 447 ms | yes | live | 2026-10-25 |
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 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:05 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.
