MiniMax M3
MiniMax's June 1 2026 open-weight flagship — the first open model to combine frontier agentic coding, native multimodality, and a 1M-token context. Built on MiniMax Sparse Attention (MSA): ~1/20th the per-token compute at 1M context, >9x faster prefill, >15x faster decode. SWE-bench Pro 59.0% surpasses GPT-5.5 and Gemini 3.1 Pro and approaches Opus 4.7; Terminal-Bench 2.1 66.0%, OSWorld-Verified 70.06%. Benchmarks were self-reported and unverified at launch (weights and technical report due within ~10 days). Launch-promo pricing was $0.30/$1.20 per 1M tokens.
1M
tokens
64K
tokens
$0.6
per 1M tokens
$2.4
per 1M tokens
80
tokens/sec
Jun 2026
2026-06-01
$1.50
per 1M tokens
59.3
quality per $
Capabilities
Benchmarks
Pricing History
Compare With
Open Source — Licensed under Open weight (modified MIT expected)
About MiniMax M3
MiniMax M3 is a open-source AI model by MiniMax, released on June 1, 2026. It supports a context window of 1000K tokens and can generate up to 64K output tokens.
At $0.6 per million input tokens and $2.4 per million output tokens, its blended cost of $1.50/1M tokens puts it in the mid-range pricing tier. Its value score of 59.3 reflects the balance of quality and cost.
MiniMax M3 is available as an open-source model under the Open weight (modified MIT expected) license, meaning you can self-host it for predictable costs or use it through API providers like Swfte Connect.
Using MiniMax M3 with Swfte
Access MiniMax M3 through Swfte Connect, our unified LLM gateway. Connect gives you a single API for 50+ models, with automatic routing, cost optimization, and fallback handling. You can also try MiniMax M3 in our AI Playground before integrating.