MiniMax Launches Open Multimodal Video Model H3 and M3.1-Flash With 1M Context
testingcatalog · x · 2026-10-19
MiniMax's API docs reveal two updates:
- MiniMax H3 is now available, a new-generation open general-purpose multimodal video model.
- New text models: MiniMax-M3.1-Flash (1M context, tunable thinking depth) and MiniMax-M3 (1M context, 100+ tps output), both positioned as frontier multimodal coding models callable via the Anthropic or OpenAI SDK.
- Legacy M2.7 (204,800 context, 60 tps, described as beginning "recursive self-improvement") and M2.5 families remain available.
More from Models
- GrokBot Cracks in Minutes What Humans Couldn't in Hours: 10-Page Report Tracing Image Spread on X — FlorianGallwitz · 2026-10-19
- antirez: New DwarfStar GGUF checkpoint for DeepSeek v4.1 Flash boosts token efficiency — antirez · 2026-10-19
- "Following the rules" got flagged as circumventing safety systems for a full workday — MoonL88537 · 2026-10-19
- Liquid's d1-3B decision model drives robot skills at 157ms on Intel Arc B580 — helloiamleonie · 2026-10-19
- Early hands-on: MiniMax-M3.1-Flash API impresses with raw speed — airesearch12 · 2026-10-19
- 5 LLM fine-tuning techniques explained visually, from 3 years of practice — usamawahabkhan · 2026-10-19