MiniMax-M3

MiniMax·China·Released 2026-06-01

Open weightsGeneral LLMReasoningTool useVision

1Mcontext

$0.3 / $1.2per 1M tokens (in / out)

427Bparameters

15stories in our feed

Specs

hover a value to see its source

Context window1M

Max output512K

Parameters427B

Knowledge cutoff—

Licenceother

Commercial useRestricted

InputText · Image · Video

OutputText

HF downloads (30d)155.9K

Familyminimax

API price

USD per 1M tokens

Input$0.3

Output$1.2

Cached input$0.06

Change history

2026-09-24 Context 1M → 1Mmodels.dev

News timeline

from the AITOP feed, one story per eventAll 15 →

2026-10-23

MiniMax M3 与 Kimi K3 更名引发生成原因猜测Teortaxes

2026-10-10

MiniMax M3 在 NVIDIA Vera Rubin NVL72 上运行MiniMax_AI

2026-10-07

SambaCloud 上 MiniMax M3 开启 Prompt Caching，输入成本降 90%SambaNova

2026-09-30

SambaNova九月动态：缓存功能上线MiniMax M3SambaNova

2026-09-22

arXiv 论文提出 opinion leader dynamics，解释稀疏注意力如何塑造 token 聚类arXiv: DeepSeek

2026-09-07

Saudi HUMAIN-M3 Builds on MiniMax M3pandaily

2026-09-02

Harness-of-Harness Framework for Autonomous CodingarXiv: DeepSeek

2026-07-22

MiniMax M3 模型在 Fireworks 平台开放训练Fireworks AI

2026-07-18

Kimi K3 在 front-end coding 上超越所有美国模型，比多数闭源模型更小shao__meng

2026-07-11

MiniMax M3内核用KV-stationary策略解决稀疏注意力内存瓶颈Fireworks AI

2026-07-07

MiniMax M3 多模态长运行智能体模型解析elvis

2026-07-01

vLLM v0.24.0 发布：支持 MiniMax-M3 和 DeepSeek-V4vLLM

2026-06-29

GLM 5.2、Opus 4.7、GPT 5.5、Minimax M3 在 QA 基准上对比Browser Use

2026-06-27

hf-claude 让 Claude Code 接入超100个开源模型，含 GLM 5.2、MiniMax-M3 等AK