Models › MiMo-V2.6 family

MiMo-V2.6-Flash-RL

XiaomiMiMo MIT out 34 days ago

311B parameters (8 of 256 experts per token), text in, text out, 1M tokens of context. 10 providers serve it from $0.10 per million input tokens.

Verified 2026-10-25 Rank #302 in LLMs · score 83.6 method v1 Sources 4 Downloads 30d 91k · #322 of 619 tracked

Changed this week: 1 new quantized build on the Hub (38 in all).2026-10-24

Can I run it?187GB at INT4Estimateneeds a multi-GPU nodemethod
What does it cost?$0.10/1M incheapest of 8Darkbloomfp42026-10-24
Can I use it?MITPermissivelicence file2026-10-24
Do people use it?91kdownloads, 30 days#322 of 619 we track · below the median+21 likes, downloads −10% this week581 likesHugging Face2026-10-24

Worth reading

Picked by rule from 17 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Latest · 28 days ago

Diagnosing and Mitigating Tool-Call Repetition in MiMo-v2.6

MiMo-V2.6 models exhibited tool-call repetition where agents issued identical calls repeatedly, wasting context and stalling tasks.

  • The flooding penalty threshold was lowered from 32 calls to eight to suppress abnormal call volume.
  • Applying the stricter flooding penalty at full scale required restarting 20 training steps at an estimated cost of $2.31 million.
mimo.xiaomi.com 2026-09-27 2 HN pts · 1 comments 1.8k words about the MiMo V2.6 family family sells inference

17 best of 17 tracked

The 1 pick above, plus these 16. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. ArticleMiMo v2.6mimo.xiaomi.com · 2026-09-21 · 1,130 HN pts · official
  2. ArticleMiMo-v2.6-Pro: Intelligence, Performance and Price Analysisartificialanalysis.ai · 2026-09-22
  3. ArticleXiaomi MiMo v2.6 Pro: New best open-weight LLM with simple designsebastianraschka.com · 2026-09-22
  4. ArticleMiMo-v2.6-Flash: on intelligence/price Pareto frontierartificialanalysis.ai · 2026-09-27 · 2 HN pts
  5. ArticleMiMo-v2.6-Flashmimo.mi.com · 2026-09-22 · 2 HN pts · official
  6. ArticleXiaomi MiMo-v2.6 Breaks Cover: A 1T-Class Chinese Lab Trains in Publicforkast.news · 2026-09-18 · 3 HN pts
  7. DiscussionMiMo-v2.6 is "simply" the besttwitter.com · 2026-09-22 · 2 HN pts
  8. ArticleMiMo-V2.6-Flash-RL: Xiaomi's Efficient 309B Omnimodal Model | MindStudiomindstudio.ai · 2026-09-26
  9. ArticleMiMo-V2.6: Scaling Reinforcement Learning from Scratch | Pratham Patelprathamp.com · 2026-09-26
  10. ArticleMiMo-V2.6: Xiaomi's RL Paper, Verified vs Reportedorcarouter.ai · 2026-09-23
  11. ArticleXiaomi MiMo-V2.6: Architecture, Benchmarks & Economics | Vantaigevantaige.io · 2026-09-23
  12. ArticleMiMo-V2.6-Flash Official Benchmarks: 30 RL Steps and Agent Results | Tabbitgo.tabbit.ai · 2026-09-22
  13. ArticleXiaomiMiMo/MiMo-V2.6-Flash-RLkinonn-bot.github.io · 2026-09-22
  14. ArticleMiMo-V2.6-Flash Review: High-Throughput Automation Workhorse, Conditional Agent | Tabbit Browsergo.tabbit.ai · 2026-09-22
  15. ArticleMiMo-V2.6-Flash-RL: 309B MoE Model, 1M Context (2026) | HokAIhokai.io · undated
  16. ArticleMiMo-V2.6 | OpenLM.aiopenlm.ai · undated

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

8 providers from $0.10

8 providers with a live price, from $0.10 to $0.14 per million input tokens; the median is $0.13. One dot per provider, cheapest in teal.

$0.010$0.10$1.00$10.00Io Net: $0.10 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashDarkbloom: $0.10 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashGMICloud: $0.12 in / $0.24 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashMakora: $0.13 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashRelace: $0.13 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashNovita: $0.14 in / $0.28 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/XiaomiMiMo/MiMo-V2.6-Flash-RL?expand[]=inferenceProviderMappingCohere: $0.14 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashXiaomi: $0.14 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashcheapest $0.10median $0.13
$0.010$0.10$1.00$10.00Io Net: $0.10 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashDarkbloom: $0.10 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashGMICloud: $0.12 in / $0.24 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashMakora: $0.13 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashRelace: $0.13 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashNovita: $0.14 in / $0.28 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/XiaomiMiMo/MiMo-V2.6-Flash-RL?expand[]=inferenceProviderMappingCohere: $0.14 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashXiaomi: $0.14 in / $0.28 out per 1M · observed 2026-10-24 · https://openrouter.ai/xiaomi/mimo-v2.6-flashcheapest $0.10 · median $0.13 (white tick)
#Cheapest 5In /1MOut /1MContextQuantSeen
1 Darkbloom $0.10 $0.28 1M fp4 Oct 24
2 Io Net $0.10 $0.28 1024K fp8 Oct 24
3 GMICloud $0.12 $0.24 1.0M bf16 Oct 24
4 Makora $0.13 $0.28 1M fp8 Oct 24
5 Relace $0.13 $0.28 1M mxfp4 Oct 24

All 10 providers →

Where it's used

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 (10 API providers); 38 quantized builds for local use.

  • As a service10Cohere, Darkbloom, DeepInfra, GMICloud +6
  • On own hardware38quantized builds
  • Built on75 fine-tunes · 2 adapters · 0 merges
  • In apps3Hugging Face Spaces

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Needs an 8×80 GB node even at INT4

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.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 187 GB of weights (estimate; KV cache and activations extra)INT4187 GBFP8: about 373 GB of weights (estimate; KV cache and activations extra)FP8373 GBBF16: about 746 GB of weights (estimate; KV cache and activations extra)BF16746 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 187 GB of weights (estimate; KV cache and activations extra)INT4187 GBFP8: about 373 GB of weights (estimate; KV cache and activations extra)FP8373 GBBF16: about 746 GB of weights (estimate; KV cache and activations extra)BF16746 GB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

Coverage peaked at launch and is still arriving

Teal is the weights (2026-09-21), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.

What people built on it

45 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

45 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

38 quantized (Hugging Face model tree)5 fine-tunes (Hugging Face model tree)2 adapters (Hugging Face model tree)38 quantized5 fine-tunes2 adapters
38 quantized (Hugging Face model tree)5 fine-tunes (Hugging Face model tree)2 adapters (Hugging Face model tree)38 quantized5 fine-tunes2 adapters

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

16 rows

Benchmarks 16 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (16)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
BenchmarkValueSetting / notesReported byDate
author-reported Agents’ Last Exam27.6MiMo-V2.6 Flashmodel author2026-10-07
author-reported AutomationBench v1.0.652.3MiMo-V2.6 Flashmodel author2026-10-07
author-reported CyberGym95.1MiMo-V2.6 Flashmodel author2026-10-07
author-reported DeepSWE · Write code with agents67.9MiMo-V2.6 Flashmodel author2026-10-07
author-reported ExploitGym6MiMo-V2.6 Flashmodel author2026-10-07
author-reported JobBench61.2MiMo-V2.6 Flashmodel author2026-10-07
author-reported MiMo Code Bench61.2MiMo-V2.6 Flashmodel author2026-10-07
author-reported MiMo Cyber Bench77.2MiMo-V2.6 Flashmodel author2026-10-07
author-reported MiMo VisualCoding71.5MiMo-V2.6 Flashmodel author2026-10-07
author-reported OSWorld-Verified80.8MiMo-V2.6 Flashmodel author2026-10-07
author-reported ProgramBench26MiMo-V2.6 Flashmodel author2026-10-07
author-reported SEC Bench Pro47.5MiMo-V2.6 Flashmodel author2026-10-07
author-reported Terminal-Bench · Write code with agents28.8MiMo-V2.6 Flashmodel author2026-10-07
author-reported Terminal-Bench · Write code with agents87.6MiMo-V2.6 Flashmodel author2026-10-07
author-reported Terminal-Bench · Write code with agents25.3MiMo-V2.6 Flashmodel author2026-10-07
author-reported Toolathlon · Call tools and APIs73.6MiMo-V2.6 Flashmodel author2026-10-07

What the card says

1 cited sentence

Summary generated, cited

The architecture supports a 1M token context length and is released under an MIT license.[1][2]

  1. “Context Length: 1M tokens”
  2. license

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.

Use-case evidence

6 use cases

What it is good at 6 use cases

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.

Facts

22 fields

Listed in LLMs · MoE

Facts

Pipeline tagtext-generation
Librarytransformers
ArchitecturesMiMoV2ForCausalLM
Model typemimo_v2
Dtypebfloat16
Params by dtypeBF16: 4,101,308,032 · F32: 12,032 · F8_E4M3: 3,859,808,256 · U8: 302,795,194,368
Size bandXXL
MoEyes
Experts256
Experts per token8
Context length1,048,576 (config.json max_position_embeddings)
Languagesen, zh
LicenseMIT (mit)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

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.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for MiMo-V2.6-Flash-RL: 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.

All variants

40 variants

Lineage

Original release (shipped quantized by the lab) Family: XiaomiMiMo/MiMo-V2.6 45 derivatives on the Hub

quantized: 38adapter: 2finetune: 5

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
ggml-org/MiMo-V2.6-Flash-RL-GGUF quantized GGUF — — mit ggml-org 19 21.2k
ProCreations/MiMo-V2.6-Flash-RL-NVFP4 quantized NVFP4 159B XXL mit ProCreations 13 14.3k
YanissAmz/MiMo-V2.6-Flash-RL-GGUF quantized GGUF — — mit YanissAmz 7 13.1k
patrickbdevaney/MiMo-V2.6-Flash-REAP50-GGUF quantized GGUF — — mit patrickbdevaney 6 7.6k
Baekpica/MiMo-V2.6-Flash-RL-Mixed-Quant-GGUF quantized -Quant — — mit Baekpica 8 6.1k
pfeifferj/MiMo-V2.6-Flash-RL-GSQ-RCO-GGUF quantized GGUF — — mit pfeifferj 9 5.5k
dealignai/MiMo-V2.6-Flash-RL-UNCENSORED quantized — 159B XXL mit dealignai 39 4k
TensorFold/MiMo-V2.6-Flash-RL-MLX-4bit-MTP quantized MLX 309B XXL mit TensorFold 11 5k
infosave/MiMo-V2.6-Flash-RL-cmf quantized — — — mit infosave 1 4.4k
benthecarman/MiMo-V2.6-Flash-RL-exl3 quantized exl3 45.8B XL mit benthecarman 9 1.8k
TrevorJS/MiMo-V2.6-Flash-RL-GGUF quantized GGUF — — mit TrevorJS 7 1.8k
mlx-community/MiMo-V2.6-Flash-RL-mxfp4-q8 quantized mxfp4 309B XXL mit mlx-community 1 2k
inferencerlabs/MiMo-V2.6-Flash-RL-MLX-Q9 quantized MLX — — — inferencerlabs 1 1.2k
Baekpica/MiMo-V2.6-Flash-RL-GGUF quantized GGUF — — mit Baekpica 1 1.1k
MorinoNushi/MiMo-V2.6-Flash-RL-Uncensored-Heretic-LoRA-GGUF adapter GGUF — — mit MorinoNushi 8 681
r0b0tlab/MiMo-V2.6-Flash-RL-NVFP4 quantized NVFP4 159B XXL mit r0b0tlab 0 1k
ProCreations/MiMo-V2.6-Flash-RL-IQ3_XXS-GGUF quantized GGUF — — mit ProCreations 4 831
patrickbdevaney/MiMo-V2.6-Flash-REAP50 quantized — 159B XXL mit patrickbdevaney 2 922
MorinoNushi/MiMo-V2.6-Flash-RL-Uncensored-Heretic-GGUF quantized GGUF — — mit MorinoNushi 8 617
tacodevs/MiMo-V2.6-Flash-RL-MLX-REAP25-mixed-GPTQ-MTP quantized MLX 233B XXL mit tacodevs 2 871
tacodevs/MiMo-V2.6-Flash-RL-MLX-REAP50-mxfp4-MTP quantized MLX 157B XXL mit tacodevs 1 897
AxionML/MiMo-V2.6-Flash-RL-NVFP4 quantized NVFP4 159B XXL mit AxionML 0 880
tiyuvta/MiMo-V2.6-Flash-RL-NVFP4 quantized NVFP4 159B XXL mit tiyuvta 3 721
tacodevs/MiMo-V2.6-Flash-RL-MLX-REAP37-mixed-GPTQ-MTP quantized MLX 195B XXL mit tacodevs 1 773
Yanun/MiMo-V2.6-Flash-RL-MXFP4 quantized MXFP4 — — mit Yanun 1 751
kernelpool/MiMo-V2.6-Flash-MXFP4-GGUF quantized MXFP4 — — mit kernelpool 2 696
nativ-community/MiMo-V2.6-Flash-RL-MLX-4bit quantized MLX 310B XXL mit nativ-community 0 761
primitive-ai/MiMo-V2.6-Flash-RL-NVFP4 quantized NVFP4 159B XXL mit primitive-ai 1 475
drowzeys/MiMo-V2.6-Flash-RL-Abliterated quantized — 311B XXL mit drowzeys 7 173
sayyidfareed/MiMo-V2.6-Flash-RL-MLX-4bit-MTP quantized MLX 309B XXL mit sayyidfareed 0 486
AlazarM/MiMo-V2.6-Flash-RL-MLX-4bit quantized MLX 310B XXL mit AlazarM 0 464
tacodevs/MiMo-V2.6-Flash-RL-MLX-REAP25-mixed3bit-GPTQ-MTP quantized MLX 233B XXL mit tacodevs 0 457
vcruz305/MiMo-V2.6-Flash-RL-dflash-EXL3-4.0bpw quantized EXL3 368M S mit vcruz305 0 390
jedisct1/MiMo-V2.6-Flash-RL-GGUF-100K quantized GGUF — — mit jedisct1 1 299
com-kotobalabs/MiMo-V2.6-Flash-Uncensored-NVFP4 finetune NVFP4 159B XXL mit com-kotobalabs 6 48
0xSojalSec/MiMo-V2.6-Flash-RL-Uncensored-Heretic-LoRA-GGUF adapter GGUF — — mit 0xSojalSec 2 168
OsaurusAI/MiMo-V2.6-Flash-RL-JANG_2L finetune — 311B XXL mit OsaurusAI 0 233
sdworld/MiMo-V2.6-Flash-SD quantized — 83.7B XL mit sdworld 1 108
vultr/MiMo-V2.6-Flash-RL-GGUF quantized GGUF — — mit vultr 0 128
dealignai/MiMo-V2.6-Flash-RL-ABLITERATED finetune — — — mit dealignai 2 0

Repository changelog

11 commits
changed in the last 7 daysOther: 2026-10-24all changes

Repository changelog 11 commits

DateCommitSHA
2026-10-09 Align preprocessor config to training values (min/max_pixels, image_mean/std) dcd3fd8
2026-09-22 Fix typo in report 5711b26
2026-09-22 Fix invalid JSON in DFlash config (#2) b2674c7
2026-09-22 Add ModelScope links 28a6a04
2026-09-21 Add assets/architecture.png 3b38d06
2026-09-21 Add README.md 8b939dc
2026-09-21 Add files using upload-large-folder tool 6fa50b3
2026-09-21 Upload MiMo_V2_6_technical_report.pdf daed50f
2026-09-21 Add files using upload-large-folder tool cf72420
2026-09-21 Add files using upload-large-folder tool 53bc18c
2026-09-21 initial commit 990f1a0

All providers

10 providers · 10 endpoints

Where to run 10 providers · 10 endpoints

Hosted endpoints observed for MiMo-V2.6-Flash-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: $0.09/M.

ProviderModel refContextQuant$/M in$/M outtok/sTTFTToolsStatusObserved
DeepInfra
openrouter
xiaomi/mimo-v2.6-flash 1M fp8 $0.09 $0.18 — — yes status -5 2026-10-24
Io Net
openrouter
xiaomi/mimo-v2.6-flash 1024K fp8 $0.10 $0.28 — — yes live 2026-10-24
Darkbloom
openrouter
xiaomi/mimo-v2.6-flash 1M fp4 $0.10 $0.28 — — yes live 2026-10-24
GMICloud
openrouter
xiaomi/mimo-v2.6-flash 1.0M bf16 $0.12 $0.24 — — yes live 2026-10-24
Makora
openrouter
xiaomi/mimo-v2.6-flash 1M fp8 $0.13 $0.28 — — yes live 2026-10-24
Relace
openrouter
xiaomi/mimo-v2.6-flash 1M mxfp4 $0.13 $0.28 — — yes live 2026-10-24
Novita
hf, openrouter
xiaomimimo/mimo-v2.6-flash 1M fp8 $0.14 $0.28 20.4 4234 ms yes live 2026-10-25
Cohere
openrouter
xiaomi/mimo-v2.6-flash 1M unknown $0.14 $0.28 — — yes live 2026-10-24
Xiaomi
openrouter
xiaomi/mimo-v2.6-flash 1M fp8 $0.14 $0.28 — — yes live 2026-10-24
Venice
openrouter
xiaomi/mimo-v2.6-flash 977K fp8 $0.17 $0.35 — — yes status -5 2026-10-24

Articles and coverage

17 pieces

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.

Official3

mimo.xiaomi.com2026-09-21officialHN front pageabout the MiMo V2.6 familyHN 1,130 pts · 483 comments3,621 words
Xiaomi releases the MiMo-V2.6 series on September 22nd, 2026, introducing MiMo-V2.6-Pro, MiMo-V2.6-Flash, and the new MiMo-V2.6-Pro-UltraSpeed variant delivering up to 20x faster output speed at the same quality. The models completed 30 RL steps over roughly 750k trajectories at costs of about $0.85M and $2.62M respectively, with DeepSWE v1.1 scores rising from 48.8 to 65.68 for the Pro model. This launch marks a key step in scaling RL compute on verifiable, complex tasks.written, reviewed, cited
  • MiMo-V2.6-Pro-UltraSpeed delivers up to 20x faster output speed at the same quality for users requiring extreme generation speed.
  • The training run completed in under six days with MiMo-V2.6-Pro costing about $2.62M and MiMo-V2.6-Flash costing about $0.85M.
mimo.xiaomi.com2026-09-27officialabout the MiMo V2.6 familyHN 2 pts · 1 comments1,793 words
MiMo-V2.6 models exhibited tool-call repetition where agents issued identical calls repeatedly, wasting context and stalling tasks. The team defined exact within-turn repetition as (N-U)/N and distinguished it from flooding, then adopted an MOPD-based approach to fix the issue. This solution cost approximately $90,000, significantly less than the $2.31 million estimated for the penalty-threshold fix, and reduced repetition rates substantially.written, reviewed, cited
  • The flooding penalty threshold was lowered from 32 calls to eight to suppress abnormal call volume.
  • Applying the stricter flooding penalty at full scale required restarting 20 training steps at an estimated cost of $2.31 million.
mimo.mi.com2026-09-22officialabout the MiMo V2.6 familyHN 2 pts · 0 comments311 words

Analysis6

artificialanalysis.ai2026-09-27authorityabout the MiMo V2.6 familyHN 2 pts · 1 comments961 words
MiMo-V2.6-Flash is a 309 billion parameter Mixture of Experts reasoning model released by Xiaomi on September 21, 2026, under the MIT license. It scores 38 on the Artificial Analysis Intelligence Index but operates at 60 tokens per second with a 4.42s time to first token. The model supports text and image input with a 1M token context window and is available through 4 API providers.written, reviewed, cited
  • The model uses 15 billion active parameters during inference despite having 309 billion total parameters.
  • MiMo-V2.6-Flash generated 240M output tokens during evaluation, which is significantly higher than the median of 140M.
forkast.news2026-09-18about the MiMo V2.6 familyHN 3 pts · 0 comments643 words
mindstudio.ai2026-09-261,399 words
prathamp.com2026-09-26about the MiMo V2.6 family8,707 words
orcarouter.ai2026-09-23about the MiMo V2.6 family
vantaige.io2026-09-23about the MiMo V2.6 family

Community1

twitter.com2026-09-22about the MiMo V2.6 familyHN 2 pts · 0 comments
All coverage tracked for this model (17) — append-only, nothing is ever removed
TitleSourcePublishedFirst seenSignals
MiMo-V2.6-Flash-RL: 309B MoE Model, 1M Context (2026) | HokAIhokai.io—2026-10-07exa
MiMo-V2.6 | OpenLM.aiopenlm.ai—2026-10-07exa
Diagnosing and Mitigating Tool-Call Repetition in MiMo-v2.6mimo.xiaomi.com official2026-09-272026-10-07hn
MiMo-v2.6-Flash: on intelligence/price Pareto frontierartificialanalysis.ai2026-09-272026-10-07hn
MiMo-V2.6-Flash-RL: Xiaomi's Efficient 309B Omnimodal Model | MindStudiomindstudio.ai2026-09-262026-10-07exa
MiMo-V2.6: Scaling Reinforcement Learning from Scratch | Pratham Patelprathamp.com2026-09-262026-10-07exa
MiMo-V2.6: Xiaomi's RL Paper, Verified vs Reportedorcarouter.ai2026-09-232026-10-07exa
Xiaomi MiMo-V2.6: Architecture, Benchmarks & Economics | Vantaigevantaige.io2026-09-232026-10-07exa
MiMo-v2.6-Flashmimo.mi.com official2026-09-222026-10-07hn
MiMo-v2.6 is "simply" the besttwitter.com2026-09-222026-10-07hn
Xiaomi MiMo v2.6 Pro: New best open-weight LLM with simple designsebastianraschka.com2026-09-222026-10-07hn
MiMo-v2.6-Pro: Intelligence, Performance and Price Analysisartificialanalysis.ai2026-09-222026-10-07hn
MiMo-V2.6-Flash Official Benchmarks: 30 RL Steps and Agent Results | Tabbitgo.tabbit.ai2026-09-222026-10-07exa
XiaomiMiMo/MiMo-V2.6-Flash-RLkinonn-bot.github.io2026-09-222026-10-07exa
MiMo-V2.6-Flash Review: High-Throughput Automation Workhorse, Conditional Agent | Tabbit Browsergo.tabbit.ai2026-09-222026-10-07exa
MiMo v2.6mimo.xiaomi.com official2026-09-212026-10-07hn
Xiaomi MiMo-v2.6 Breaks Cover: A 1T-Class Chinese Lab Trains in Publicforkast.news2026-09-182026-10-07hn

Similar models

3 models

Similar models

  • DeepSeek-V4-Flash-0731 · 304B
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
  • DeepSeek-V4-Flash · 291B
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
  • Hy3 · 299B
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license

Sources

4 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:05 UTC · Last verified 2026-10-25

Follow MiMo-V2.6-Flash-RL

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.