MiniMax M3
Open-weight 1M-context model with native multimodality and computer use.
About
MiniMax's June 2026 release pairs a 1M-token context with native image and video understanding plus direct computer control, built on the new MiniMax Sparse Attention (MSA) architecture. The team reports MSA cuts per-token compute roughly 20x at 1M context, giving 9x faster prefill and 15x faster decoding than the previous M2 generation.
Benchmark figures (59.0% SWE-Bench Pro, 66.0% Terminal-Bench 2.1) come from MiniMax's own testing and are not yet independently confirmed.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- swe-bench-pro: 59.0 (MiniMax-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Released June 1, 2026; weights ~June 11; MiniMax Sparse Attention (MSA) architecture; 1M context, native image+video understanding, computer use in one model; SWE-Bench Pro 59.0% and Terminal-Bench 66.0% are MiniMax-reported; independent verification pending. License: Open-weight under a custom MiniMax license (not OSI-approved); confirm terms before commercial use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Released June 1, 2026; weights ~June 11; MiniMax Sparse Attention (MSA) architecture
- · 1M context, native image+video understanding, computer use in one model
- · SWE-Bench Pro 59.0% and Terminal-Bench 66.0% are MiniMax-reported; independent verification pending