MiniMax M3 Launches as First Open-Weights Model Combining Frontier Coding, 1M Context, and Native Multimodality
VentureBeat·medium signal
MiniMax M3 (June 1, Shanghai) posts 59.0% on SWE-Bench Pro, 66.0% on Terminal-Bench 2.1, and 83.5 on BrowseComp — edging GPT-5.5 but below Claude Opus 4.8 — at roughly 5–10% of frontier cost ($0.30/$1.20 per 1M on a launch promo). Major caveat: all figures are vendor-run on MiniMax's own infra, and open weights had not shipped at launch, promised on Hugging Face/GitHub within ten days. Watch for the actual weight drop before trusting the benchmarks.