MiniMax M3 Becomes the First Open-Weights Model to Combine Million-Token Context, SWE-Bench Dominance, and Native Multimodality
The Chinese lab's M3 model hits 59% on SWE-Bench Pro with a million-token context window and native multimodal support — a combination no open-weights model has achieved before, and a sign that the frontier gap is narrowing fast.
MiniMax released M3 over the weekend, and the benchmarks are hard to dismiss. As @CollinearAI documented, M3 is "the first open-weights model to combine three frontier capabilities": a one-million-token context window, a 59.0% score on SWE-Bench Pro, and native multimodal processing. Each of those capabilities individually would have been frontier-only territory twelve months ago. Bundled together in an open-weights release, they represent a genuine inflection point for what's available outside the walled gardens of OpenAI and Anthropic.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.