models · minimax

MiniMax M2.7

MiniMax M2.7. 204.8K context, 199 tok/s. Arena 1181. Latest generation MiniMax model with improved instruction following and reasoning.

Compare Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.

Weight file SHA-256	7f54e26676152327d2cae77b05c8a642b70ee4c5a8ec49ede5cb254fbbd047d6
Method	hf-published-sha256

Re-check this yourself → Source → Full record →

What is recorded

Publisher	minimax
Origin	Recorded as a covered nation under 10 U.S.C. § 4872(f) — the PRC, Russia, Iran or North Korea. The registry did not record which.
Licence	MiniMax-Open
Publisher’s own licence statement	other
Parameters	228,689,764,864 (228.7B)
Context window	204.8K
Modality	text->text
Architecture	minimax_m2
Layers	62
Hidden size	3072
Attention	GQA (48 heads, 8 KV)
Native precision	F8_E4M3
Pulls recorded here	270,015
Downloads reported upstream	1,366,935

Reported scores

GPQA Diamond	87.4
IFBench	75.7
Humanity's Last Exam	29.6

Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Reported to local-ai-registry

Hardware	Engine	Build measured	Decode, 1 stream	Label	Reported by
NVIDIA DGX Spark GB10	llama.cpp	UD-IQ4_XS	31.0 tok/s @ 32,768	candidate	localmaxxing
Intel Arc Pro B70 ×4 devices	llama.cpp	UD-IQ4_XS	17.7 tok/s @ 2,048	candidate	localmaxxing
Ryzen AI Max+ 395	llama.cpp	UD-Q3_K_M	13.9 tok/s @ 16,384	candidate	localmaxxing
Measured outside 2K–32K context — decode speed depends heavily on context, so these are not comparable with the rows above
NVIDIA DGX Spark GB10	llama.cpp	UD-IQ4_XS	30.9 tok/s @ 108,000	candidate	localmaxxing
NVIDIA DGX Spark GB10	vllm	AWQ-4bit	27.8 tok/s @ 190,000	candidate	localmaxxing
Intel Arc Pro B70 ×4 devices	llama.cpp	UD-IQ4_XS	14.4 tok/s @ 64	candidate	localmaxxing

From local-ai-registry (MIT), commit 124959d: 21 runs across 6 machines, 0 validated — model revision and runtime pinned, launch accepted — and 21 candidate, their label for useful evidence without a reproducible-launch promise.

Elsewhere in the registry

All models → Browse with filters → What will run on my hardware? →

Registry record as of 2026-08-13. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.