Models/MiniMax/MiniMax M2

MiniMax M2

MiniMax-M2 is a compact, high-efficiency large language model optimized for end-to-end coding and agentic workflows. With 10 billion activated parameters (230 billion total), it delivers near-frontier intelligence across general reasoning,...

01

Identity

204,800Context in
131,072Context out
2025-10-23 (OpenRouter created 1761252093)Released
OpenWeights
Canonical name
MiniMax M2
Aliases
minimax/minimax-m2, MiniMaxAI/MiniMax-M2
Continuum hosted id
Not listed as a Continuum hosted id
OpenRouter slug
minimax/minimax-m2
Hugging Face
MiniMaxAI/MiniMax-M2
Modalities
text
Open weights Not on Continuum host list Open

Context window on the 2026-08-19 OpenRouter row: 204,800 tokens.

Max completion tokens on that row: 131,072.

Input / output on that row: $0.26 / $1.02 per 1M tokens.

Architecture fields: text->text · Other.

Hugging Face id on that row: MiniMaxAI/MiniMax-M2.

02

Should I use this for coding agents

MiniMax-M2 is a compact, high-efficiency large language model optimized for end-to-end coding and agentic workflows.

When you want the 2025-10-23 catalog SKU, not a later rename.

DeepSWE

No official DeepSWE mini-swe-agent row for this identity on the public board we fetched.

omitted, not invented as of 2026-08-13 source
Artificial Analysispublished integer
29
bench
Artificial Analysis
version
Intelligence Index
harness
published integer
metric
Intelligence Index
value
29
independent as of 2026-08-19 source
Vals Index

We did not open a vals.ai card for this identity. Chip omitted.

omitted, not invented as of 2026-08-19 source

Ranked view: Best coding models: the independent leaderboard puts this row and every other card that carries an independent coding score on one board, with the confidence intervals left visible.

Named benches, printed only

Copied from AA or vals HTML we opened. 0.0% placeholder bars are omitted. Do not average these into the hero tiles, and do not invent CI, $/task, or steps for Artificial Analysis.

BenchPrintedNoteSourceURLAs of
SciCode 36.1% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
Humanity's Last Exam 13.7% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
GPQA Diamond 77.7% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
CritPt 0.9% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
AA-Omniscience Accuracy 23.2% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
AA-LCR 65% Printed on the AA Intelligence Evaluations grid for MiniMax-M2. Not a DeepSWE chip. Artificial Analysis MiniMax-M2 source 2026-08-19
03

When not to use it

Trust this before you buy

  • When you need a Continuum-hosted SKU from this lab, use the hosted card on this hub instead of this catalog row.
  • $0.26 in / $1.02 out per 1M on the 2026-08-19 OpenRouter row. Context 204,800 in / 131,072 out.
  • Need a denser published sibling on this hub? MiniMax M3 is the card already on the cluster.
Get Plus See plans

Plus is $25/mo with $25 weekly hosted usage. The Mac app stays free with your own keys. Get Plus is not Download for Mac.

04

Artifact and Hugging Face downloads

Artifact

SPDX / license
other
Total params
HF safetensors.total: 228,689,764,864 stored tensors
Native precision
HF tensors F32 + BF16 + F8_E4M3
Repo size
HF API usedStorage 230,135,473,737 bytes
HF created
2025-10-22T13:45:10.000Z
Official HF repo
MiniMaxAI/MiniMax-M2
HF downloads
222,048
HF likes
1,502

Official weights id on the 2026-08-19 OpenRouter row: MiniMaxAI/MiniMax-M2. Community GGUF is a quant, not a second Continuum card.

Official weights

Use this model on Hugging Face Files on Hugging Face

This is the model repo, not Download for Mac. Downloads and likes from the Hugging Face API on 2026-08-19.

Hub file tree

Safetensor shards
130
Shard bytes
230,134,264,792 bytes
Other file bytes
28,325,547 bytes
Tree file count
153
Hub usedStorage
230,135,473,737 bytes

usedStorage and the sum of .safetensors sizes can differ (LFS pointers, non-shard files, duplicate copies). Cited separately. Source: recursive tree API, 2026-08-19.

PathBytes
model-00000-of-00130.safetensors3,693,062,776
model-00001-of-00130.safetensors1,208,321,208
model-00002-of-00130.safetensors2,463,868,968
model-00003-of-00130.safetensors1,208,321,208
model-00004-of-00130.safetensors2,463,868,968
model-00005-of-00130.safetensors1,208,321,208
model-00006-of-00130.safetensors2,463,868,968
model-00007-of-00130.safetensors1,208,321,208
model-00008-of-00130.safetensors2,463,868,968
model-00009-of-00130.safetensors1,208,321,208
model-00010-of-00130.safetensors2,463,868,968
model-00011-of-00130.safetensors1,208,321,208
model-00012-of-00130.safetensors2,463,868,968
model-00013-of-00130.safetensors1,208,321,208
model-00014-of-00130.safetensors2,463,868,968
model-00015-of-00130.safetensors1,208,321,208
model-00016-of-00130.safetensors2,463,868,968
model-00017-of-00130.safetensors1,208,321,208
model-00018-of-00130.safetensors2,463,868,968
model-00019-of-00130.safetensors1,208,321,208
model-00020-of-00130.safetensors2,463,870,000
model-00021-of-00130.safetensors1,208,321,720
model-00022-of-00130.safetensors2,463,870,000
model-00023-of-00130.safetensors1,208,321,720
model-00024-of-00130.safetensors2,463,870,000
model-00025-of-00130.safetensors1,208,321,720
model-00026-of-00130.safetensors2,463,870,000
model-00027-of-00130.safetensors1,208,321,720
model-00028-of-00130.safetensors2,463,870,000
model-00029-of-00130.safetensors1,208,321,720
model-00030-of-00130.safetensors2,463,870,000
model-00031-of-00130.safetensors1,208,321,720
model-00032-of-00130.safetensors2,463,870,000
model-00033-of-00130.safetensors1,208,321,720
model-00034-of-00130.safetensors2,463,870,000
model-00035-of-00130.safetensors1,208,321,720
model-00036-of-00130.safetensors2,463,870,000
model-00037-of-00130.safetensors1,208,321,720
model-00038-of-00130.safetensors2,463,870,000
model-00039-of-00130.safetensors1,208,321,720
model-00040-of-00130.safetensors2,463,870,000
model-00041-of-00130.safetensors1,208,321,720
model-00042-of-00130.safetensors2,463,870,000
model-00043-of-00130.safetensors1,208,321,720
model-00044-of-00130.safetensors2,463,870,000
model-00045-of-00130.safetensors1,208,321,720
model-00046-of-00130.safetensors2,463,870,000
model-00047-of-00130.safetensors1,208,321,720
model-00048-of-00130.safetensors2,463,870,000
model-00049-of-00130.safetensors1,208,321,720
model-00050-of-00130.safetensors2,463,870,000
model-00051-of-00130.safetensors1,208,321,720
model-00052-of-00130.safetensors2,463,870,000
model-00053-of-00130.safetensors1,208,321,720
model-00054-of-00130.safetensors2,463,870,000
model-00055-of-00130.safetensors1,208,321,720
model-00056-of-00130.safetensors2,463,870,000
model-00057-of-00130.safetensors1,208,321,720
model-00058-of-00130.safetensors2,463,870,000
model-00059-of-00130.safetensors1,208,321,720
model-00060-of-00130.safetensors2,463,870,000
model-00061-of-00130.safetensors1,208,321,720
model-00062-of-00130.safetensors2,463,870,000
model-00063-of-00130.safetensors1,208,321,720
model.safetensors.index.json10,213,486
tokenizer.json9,730,160
vocab.json4,705,413
merges.txt2,414,077
figures/Bench.png1,046,882
figures/wechat.jpeg75,475
modeling_minimax_m2.py30,955
README.md22,046
docs/tool_calling_guide_cn.md16,570
docs/tool_calling_guide.md16,519
tokenizer_config.json10,893
configuration_minimax_m2.py10,158
chat_template.jinja6,780
docs/vllm_deploy_guide.md4,121
docs/vllm_deploy_guide_cn.md3,883
docs/sglang_deploy_guide.md3,652
docs/sglang_deploy_guide_cn.md3,438
docs/transformers_deploy_guide.md2,741
docs/transformers_deploy_guide_cn.md2,668
config.json2,055
docs/mlx_deploy_guide.md1,836
.gitattributes1,573
generation_config.json166

Showing the first 64 of 130 safetensor shards, plus 23 other files. The snapshot was capped at 64 shards when the tree API was read on 2026-08-19, so the remaining 66 are on the Files tab, not on this page. Totals above are the full tree API sum, not this slice. Byte sizes are Hub tree API size fields, never invented.

05

Continuum serving

OpenRouter slug minimax/minimax-m2 on the 2026-08-19 catalog.

First party: https://platform.minimax.io/.

Not on the Continuum host list fetched 2026-08-19.

Continuum hosted id not on the host list
MiniMax M2 first partyOpenRouter listing

We list first-party and OpenRouter, plus Continuum only when the live public allowlist named the id. This is not a 15-host routing table.

06

Economics

OpenRouter 2026-08-19: $0.26 in / $1.02 out per 1M tokens. Context 204,800 in / 131,072 out.

No separate internal-reasoning price on that row.

Price this model Opens the pricing calculator preloaded with MiniMax M2.
07

Provenance

ClaimSourceAs of
OpenRouter id minimax/minimax-m2; context 204800; created 1761252093; $0.26 / $1.02 per 1MOpenRouter /api/v1/models2026-08-19
AA Intelligence Index 29; 6 printed Index benchesArtificial Analysis model page2026-08-19
OpenRouter id minimax/minimax-m2, context 204,800OpenRouter /api/v1/models2026-08-19
HF downloads 222,048Hugging Face API MiniMaxAI/MiniMax-M22026-08-19
HF safetensors.total: 228,689,764,864 stored tensors; HF API usedStorage 230,135,473,737 bytes; created 2025-10-22T13:45:10.000Z; HF tensors F32 + BF16 + F8_E4M3Hugging Face API MiniMaxAI/MiniMax-M22026-08-19
130 safetensor shards; shard bytes 230,134,264,792; tree files 153Hugging Face tree API MiniMaxAI/MiniMax-M22026-08-19
08

Compare, FAQ, and Get Plus

No other card in this cluster shares a published DeepSWE official row we can put next to this one. We will not compare on vendor-blog numbers.

Continue through MiniMax's model family with MiniMax M1.

FAQ

Why does the lab blog disagree with DeepSWE or Scale?

Lab posts pick a harness, an effort, a split, and sometimes a private eval set. DeepSWE publishes the official mini-swe-agent row with cost, tokens, and steps. Scale SWE-bench Pro only counts when the public shared-harness board has a row we can fetch. A higher lab number is usually a different test, not a better one.

OpenAI-compatible call

No Continuum hosted id is verified for this model, so this page does not invent a curl target. Use the first-party API or the OpenRouter slug minimax/minimax-m2.

Get Plus See plans

Plus is $25/mo with $25 weekly hosted usage. The Mac app stays free with your own keys. Get Plus is not Download for Mac.