EventVisual AI1 source
AI Lab
MiniMax News
Official MiniMax announcements — model releases, product launches and research, each one summarized with every source covering it, by AIBriefs. RSS
LaunchAI Models7 sources
MiniMax M3.1 Flash cuts agentic task costs 70%, priced at $0.15/$0.60 per million tokens
MiniMax released M3.1 Flash on October 19, 2026, claiming 70% lower agentic task cost and nearly 3x faster online inference than M3. It's priced at $0.15/M input and $0.60/M output tokens — half M3's discounted rate. Together AI says it beats Opus 5.5 on agentic automation at ~27x lower cost, with a 1M context window and vision support.
LaunchVisual AI1 source
H3 video model gains 8-step reference-conditioned video and audio
LaunchDevelopers1 source
Inferact gains Day 0 support from MiniMax
LaunchDevelopers1 source
TokenSpeed adds Day-0 support for MiniMax models
AnalysisVisual AI1 source
H3 video model ported to TPUs for faster inference
LaunchAI Models15 sources
MiniMax releases M3.1 Flash: 428B MoE agentic model at $0.15/M input
The open-weight multimodal MoE has ~23B active parameters and a 1M-token context, with reasoning always on across five effort levels. MiniMax claims 70% lower agent costs and leads Automation-Bench v1.0.6 at 62.2, but scores 38.1 on Terminal Bench 4.0 versus 51.8 for Opus 5.
EventVisual AI2 sources
MiniMax hosts AI video production panel at Advertising Week New York
LaunchVisual AI1 source
MiniMax's FastH3-Omni generates video with synced audio on Reactor
EventBusiness14 sources
Hugging Face kicks off Open Source AI Week with Open Together party
EventBusiness1 source
MiniMax's Linda Sheng speaks at DBS Asia Future Forward Conference
LaunchAI Models1 source
MiniMax M3 model runs on NVIDIA Vera Rubin NVL72 with vLLM
AnalysisAI Models2 sources
MiniMax H3 video world model demonstrates physics capabilities
LaunchVisual AI1 source
Silverside utilizes MiniMax Music and H3 for animation and lip sync
EventBusiness2 sources
MiniMax hosts AI events at SF Tech Week
EventAI Models1 source
MiniMax to present on model quantization and speed at AI Engineer NYC
LaunchVisual AI5 sources