MiniMax M3
MiniMax M3Rankingin thinking mode?
Alternatives?
CheaperClaude Haiku 5.5unstated+3%-62%$0.10 / $0.50Remove sampling parameters+5BetterClaude Opus 5.5unstated+5%+1,424%$4 / $20No forced tool use+5
OpenGLM 5.3 Flashunstated+2%-55%$0.15 / $0.50Defaults to max · Self-hosted+4
CheaperMiMo v2.6 FlashunstatedOpen-2%-67%$0.14 / $0.28Pass back reasoning_content+3
CheaperGPT-5.6 Lunaunstated-3%-14%$0.20 / $1.20Claude Haiku 5.5unstated+3%-62%$0.10 / $0.50Remove sampling parameters+5GLM 5.3 FlashunstatedOpen+2%-55%$0.15 / $0.50Defaults to max · Self-hosted+4MiMo v2.6 FlashunstatedOpen-2%-67%$0.14 / $0.28Pass back reasoning_content+3GPT-5.6 Lunaunstated-3%-14%$0.20 / $1.20GPT-6 Lunaunstated-4%-62%$0.10 / $0.50Function calling only with none · Chat Completions+3Claude Opus 5.5unstated+5%+1,424%$4 / $20No forced tool use+5Claude Fable 5.1unstated+5%+3,710%$10 / $50Replace forced tool_choice+5Claude Sonnet 5.5unstated+4%+662%$2 / $10No forced tool use+5Claude Opus 5unstated+4%+1,805%$5 / $25
Muse Spark 1.2unstated+3%+281%$1.25 / $4.25Only auto tool_choice is supported+1GLM 5.3 FlashunstatedOpen+2%-55%$0.15 / $0.50Defaults to max · Self-hosted+4GLM 5.3unstatedOpen+2%+310%$1.40 / $4.40Memory (4-bit): 256 GB → 452 GB+5MiMo v2.6 ProunstatedOpen+1%+4%$0.43 / $0.87Memory (4-bit): 256 GB → 615 GB+4TInklingunstatedOpen-2%+229%$0.95 / $4.05Context: 524K → 256K+1MiMo v2.6 FlashunstatedOpen-2%-67%$0.14 / $0.28Pass back reasoning_content+3MiniMax M2.1unstatedOpen-7%—$0.30 / $1.20Context: 524K → 205K+1MiniMax M2.7unstatedOpen-8%—$0.30 / $1.20Context: 524K → 205K+1MiniMax M3.1 Flashthinkingn/a—$0.22Complete model response must be appendedMiniMax M2.5thinkingOpenn/a—$0.30 / $1.20Context: 524K → 205K+1MiniMax M2thinkingOpenn/a—$0.30 / $1.20Context: 524K → 205K+1MiniMax M1 40kOpenn/a——Memory (4-bit): 256 GB → 274 GB+1MiniMax M1 80kOpenn/a—$0.55 / $2.20Memory (4-bit): 256 GB → 274 GB+1MiniMax M1unstatedOpenn/a—$0.40 / $2.20Memory (4-bit): 256 GB → 274 GB+1All MiniMax models →Run it yourself
estimateHistory
6 versionsBenchmarks
all categoriesWatch out 1
Deployment status experimental · Self-hosted›
The documented MiniMax M3 baseline uses an experimental SGLang development image; pin its model revision and image digest and validate your workload. Source ↗
FAQ
How much does MiniMax M3 cost?›
$0.30 per 1M input tokens and $1.20 per 1M output tokens; cached input $0.06 per 1M; cheapest via CoreWeave at $0.23 / $0.96. A typical benchmark task costs about $0.51 at default effort.
Is MiniMax M3 free to use?›
There is no free API; the cheapest paid price is $0.23 per 1M input tokens. The weights are open, so you can run it yourself.
How good is MiniMax M3, and what is it best at?›
#51 of 336 in Chat & Knowledge, #51 of 134 in Coding, #60 of 135 in Agents & Tool Use, #60 of 197 in Writing & Language, #30 of 125 in Vision & Documents. Strongest at: Tool calling, Reasoning, Science, Long context, Knowledge, Instructions, Vision.
Is MiniMax M3 better than MiniMax M2.7?›
On Vals MedScribe, MiniMax M3 scores 87.3 vs 79.9 — higher.
How can I use MiniMax M3?›
Through the API of AtlasCloud, CoreWeave, DeepInfra, Featherless, Mara and 7 more.
What is the context window of MiniMax M3?›
524K tokens, with up to 512K output tokens.
When was MiniMax M3 released? Is there a newer version?›
Released 2026-06-01. It is the current version in its line.
Is MiniMax M3 open source?›
Yes, the weights are open (minimax-community licence).
Can I run MiniMax M3 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can MiniMax M3 read images?›
It accepts text and images, video as input.
What is MiniMax M3?›
MiniMax M3 is an open-weights language model from MiniMax, released Jun 2026, the successor to MiniMax M2.7.
What are the best alternatives to MiniMax M3?›
Claude Haiku 5.5 (cheaper), Claude Opus 5.5 (better), GLM 5.3 Flash (open weights).