Up to date
XiaomiMiMo MIT out 6 months ago
311B parameters (8 of 256 experts per token), text in, text out, 1M tokens of context. 3 providers serve it from $0.14 per million input tokens.
Verified 2026-10-26 07:27 UTC Rank #370 in LLMs · score 80.5 method v1 Sources 4 Downloads 30d 136.9k · #277 of 619 tracked
Changed this week: 2 providers no longer listed: Featherless, Novita.2026-10-23
Picked by rule from 17 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Model IDs for MiMo-V2.5-Pro and MiMo-V2.5 stopped working on Oct 21, 2026, at 10:00 (UTC+8).
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 4 picks above, plus these 13. 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.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | DeepInfra | $0.14 | $0.28 | 256K | — | Oct 26 |
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 3 API providers; 29 quantized builds for local use.
Weights only, an estimate: leave room for the KV cache. Method
About 187 GB at INT4, 373 GB at FP8, 746 GB 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.
30 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.
XiaomiMiMo/MiMo-V2.5 is a 311B parameter sparse Mixture of Experts model designed for text, image, video, and audio understanding.[1][2][3] The architecture utilizes 8 experts per token and supports a context window of up to 1M tokens.[4][5]
Generated 2026-10-07 15:03 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: 4,052,024,960 · F32: 49,295,104 · F8_E4M3: 306,655,002,624 |
| Size band | XXL |
| MoE | yes |
| Experts | 256 |
| Experts per token | 8 |
| Context length | 1,048,576 (config.json max_position_embeddings) |
| Languages | en, zh |
| License | MIT (mit) |
| Hub snapshot | 2026-10-26 |
vLLM: supported · MiMoV2ForCausalLM · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · MiMoV2ForCausalLM · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f
vLLM registry (4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60) · SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · 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: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 746 GB · Cluster (≥640 GB) |
|---|---|
| FP8 (1 byte/param) | 373 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 187 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: 29adapter: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| lukealonso/MiMo-V2.5-NVFP4 | quantized | NVFP4 | 179B | XXL | mit | lukealonso | 29 | 85.9k |
| unsloth/MiMo-V2.5-GGUF | quantized | GGUF | — | — | mit | unsloth | 51 | 33.5k |
| QuantPasture/MiMo-V2.5-GGUF | quantized | GGUF | — | — | — | QuantPasture | 44 | 22.9k |
| bartowski/MiMo-V2.5-GGUF | quantized | GGUF | — | — | mit | bartowski | 14 | 18.8k |
| spectator2026/MiMo-V2.5-AWQ-int4 | quantized | AWQ | 52.9B | XL | mit | spectator2026 | 5 | 11.5k |
| shadowlilac/MiMo-V2.5-NVFP4 | quantized | NVFP4 | 178B | XXL | — | shadowlilac | 1 | 9.2k |
| huihui-ai/Huihui-MiMo-V2.5-abliterated-GGUF | quantized | GGUF | — | — | mit | huihui-ai | 31 | 6.3k |
| jedisct1/MiMo-V2.5-coder-Q2 | quantized | GGUF | — | — | mit | jedisct1 | 77 | 3.2k |
| tnhnyzc/MiMO-V2.5-MTP-GGUF | quantized | GGUF | — | — | mit | tnhnyzc | 8 | 5.9k |
| gghfez/MiMo-V2.5-control-vectors | adapter | GGUF | — | — | cc-by-nc-4.0 | gghfez | 3 | 2.2k |
| jedisct1/MiMo-V2.5-coder-Q2-v2 | quantized | GGUF | — | — | mit | jedisct1 | 12 | 1.3k |
| gghfez/MiMo-V2.5-ikllama-GGUF | quantized | GGUF | — | — | — | gghfez | 4 | 1.7k |
| nwzjk/MiMo-V2.5-AWQ-int4 | quantized | AWQ | 52.9B | XL | mit | nwzjk | 1 | 1.5k |
| bearzi/MiMo-V2.5-MLX | quantized | MLX | — | — | mit | bearzi | 0 | 1.5k |
| dawncr0w/MiMo-V2.5-oQ4-MLX | quantized | MLX | 311B | XXL | mit | dawncr0w | 1 | 1.4k |
| DevQuasar/XiaomiMiMo.MiMo-V2.5-GGUF | quantized | GGUF | — | — | — | DevQuasar | 0 | 1.4k |
| unsloth/MiMo-V2.5 | quantized | — | 311B | XXL | mit | unsloth | 1 | 1.2k |
| Sciguy429/MiMo-V2.5-IK-GGUF | quantized | GGUF | — | — | — | Sciguy429 | 2 | 947 |
| jedisct1/MiMo-V2.5-coder-Q2-v2-MTP | quantized | GGUF | — | — | mit | jedisct1 | 6 | 663 |
| christopher-kapic/MiMo-V2.5-ROCmFP4-GGUF | quantized | GGUF | — | — | — | christopher-kapic | 3 | 618 |
| shadowlilac/MiMo-V2.5-AWQ-int4 | quantized | AWQ | 52.9B | XL | mit | shadowlilac | 0 | 662 |
| hell0ks/MiMo-V2.5-GGUF-ko | quantized | GGUF | — | — | mit | hell0ks | 3 | 442 |
| mradermacher/MiMo-V2.5-i1-GGUF | quantized | GGUF | — | — | mit | mradermacher | 2 | 455 |
| fivetech/MiMo-V2.5-colibri-peng-int4 | quantized | int4 | — | — | apache-2.0 | fivetech | 2 | 435 |
| mradermacher/MiMo-V2.5-GGUF | quantized | GGUF | — | — | mit | mradermacher | 0 | 527 |
| Niustron/Huihui-MiMo-V2.5-abliterated-GGUF | quantized | GGUF | — | — | mit | Niustron | 1 | 367 |
| Blightbow/MiMo-V2.5-Text-Only-oQ4-MLX | quantized | MLX | 309B | XXL | mit | Blightbow | 0 | 330 |
| PNC-user/MiMo_V2.5_coder_Q2-pnc | quantized | GGUF | — | — | mit | PNC-user | 0 | 200 |
| Akicou/MiMo-V2.5-NE128-72-OP | quantized | — | 159B | XXL | mit | Akicou | 1 | 98 |
| fivetech/MiMo-V2.5-colibri-peng-int4-zstd | quantized | int4 | — | — | apache-2.0 | fivetech | 1 | 92 |
| Date | Commit | SHA |
|---|---|---|
| 2026-07-09 | Update README.md | 6365158 |
| 2026-07-09 | Update README.md | 39b8981 |
| 2026-05-08 | Update tokenizer_config.json | 2fd4f89 |
| 2026-05-08 | Update README.md | ff93ba9 |
| 2026-05-08 | Upload assets/wechat.jpg (#15) | 555948a |
| 2026-05-08 | remove qr code at root path | 0c687b7 |
| 2026-05-08 | update wechat qr code and discord invite link (#14) | b8aed10 |
| 2026-05-07 | Update generation_config.json | 13b5e3f |
| 2026-04-29 | Update README.md | b095f78 |
| 2026-04-29 | Update README.md | 7d93af8 |
| 2026-04-28 | Update MiMo-V2.5 config | 4da2748 |
| 2026-04-28 | Add ModelScope url | b944074 |
| 2026-04-28 | Upload model.safetensors.index.json | e4bd52a |
| 2026-04-27 | Upload 6 files | c3f3abb |
| 2026-04-27 | Upload 5 files | 6acfb7c |
| 2026-04-27 | Upload 15 files | 4f78a4d |
| 2026-04-27 | Add files using upload-large-folder tool | c8dea8d |
| 2026-04-27 | initial commit | e6a82e2 |
Hosted endpoints observed for MiMo-V2.5, 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: $0.12/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| GMICloud openrouter | xiaomi/mimo-v2.5 | 1.0M | fp8 | $0.12 | $0.24 | — | — | yes | status -5 | 2026-10-26 |
| DeepInfra hf | XiaomiMiMo/MiMo-V2.5 | 256K | — | $0.14 | $0.28 | 9.7 | 2705 ms | yes | live | 2026-10-26 |
| Venice openrouter | xiaomi/mimo-v2.5 | 977K | fp8 | $0.40 | $2.00 | — | — | yes | status -2 | 2026-10-26 |
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: 2026-10-26 · HF API: 2026-10-26 07:17 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:03 UTC · Last verified 2026-10-26 07:27 UTC
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
