MiniMax M3 minimax-m3

minimax · minimax reasoning flagship agentic

#6 overall

params
427B
arch
moe
context
1024k
license
other open weights
released
Jun 2026
reasoning
yes
downloads/30d
176.3k

# architecture

attn shared · reasoning GQA 64q/4kv router top-4 of 128 +1 shared expert ×128 expert out ×60 layers ctx 1,048,576 427B pool
layers
60
d_model
6144
heads
64q / 4kv
head dim
128
experts
top-4 of 128 + 1 shared
vocab
200064
family
minimax_m3_vl

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# why ranked

Overall: #6 — score 91.7 ● high — all signals present

signalweightinput (0–100)
aa_intelligence 0.70 90.7
bench_composite 0.30 93.9

benchmark panel evidence:

complete panelscore (0–100)
browsecomp-v1100.0
livebench-v187.7

Coding: provisional — score 77.2 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_coding 0.30 77.2
aider_polyglot 0.30
swe_bench_verified 0.40

missing signals are dropped and the remaining weights renormalized — never imputed.

full methodology

# trend

OGM score · last 55 days
OGM score over 55 days
overall rank (up is better)
Overall rank over 55 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v5.

# benchmarks

benchmarkscoresourcedate
AA Coding Index 58.6 Artificial Analysis
AA Intelligence Index 35.7 Artificial Analysis
LMArena Elo (Code) 1487.0 LMArena Sep 2026
LiveBench Agentic Coding 40.7 LiveBench Jun 2026
LiveBench Coding 68.2 LiveBench Jun 2026
LiveBench Data Analysis 76.2 LiveBench Jun 2026
LiveBench IF 57.5 LiveBench Jun 2026
LiveBench Language 76.8 LiveBench Jun 2026
LiveBench Mathematics 76.9 LiveBench Jun 2026
LiveBench Reasoning 74.5 LiveBench Jun 2026
Apex-agents 0.3 / 1 LLM Stats
Bankertoolbench 0.8 / 1 LLM Stats
Browsecomp 0.8 / 1 LLM Stats
Cl-bench 0.2 / 1 LLM Stats
Claw-eval 0.7 / 1 LLM Stats
Imo-2025 35.0 / 42 LLM Stats
Usamo-2026 36.0 / 42 LLM Stats
Yc-bench 2100000.0 / 10000000 LLM Stats
LMArena Elo (Vision) 1240.0 LMArena Aug 2026

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
Featherless via hfrouter
Nebius nebius
SiliconFlow siliconflow
CoreWeavefp4 256k $0.23$0.96$0.05 via openrouter 99.9%
Wandb 256k $0.23$0.96 via litellm
DeepInfrafp8 512k $0.28$1.10 deepinfra 99.6%
AtlasCloudfp8 524k $0.30$1.20 atlascloud 99.9%
Fireworks 500k $0.30$1.20 via hfrouter 73
Minimaxfp8 512k $0.30$1.20$0.06 via openrouter 99.7%
Novitafp8 1000k $0.30$1.20 novita 99.7%
Parasailfp8 1024k $0.30$1.20$0.06 via openrouter 99.5%
StreamLakefp8 1000k $0.30$1.20$0.06 via openrouter 99.1%
Tencent 1000k $0.30$1.20 via litellm
Togetherunknown 512k $0.30$1.20$0.06 via openrouter 99.9%
Venicefp8 512k $0.30$1.20 venice 100.0%
GMICloudfp8 1024k $0.60$2.40$0.12 via openrouter 99.4%
SambaNovaunknown 1024k $0.60$2.40 sambanova 98.7%
ModelRunfp4 1024k $0.75$3.00$0.15 via openrouter 99.2%

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

source aliases
aa
minimax-m3
arena
MiniMax-M3
openrouter
minimax/minimax-m3