A Chinese developer of multimodal models. The video brand Hailuo and its character chat app are also both products of this company. It offers the coding assistance tool MiniMax Code and the video generation model H3. In August 2026, MiniMax Code added browser automation and goal mode features. That same month, MiniMax Code also unveiled support for agent plugin release dates. In August 2026, the company introduced a music model that generates a complete five-minute song from lyrics alone. That same month, Luma AI integrated MiniMax's H3 video model into Luma Agents.
Current rank (1M)
24
Rank over the last 30 days
2026.08.20 – 2026.09.18 · High 15 · Low 24 · Now 24
Anthropic published a 154-page report exposing unauthorized distillation by seven Chinese AI labs. But as the company itself admits, distillation is legal — the real problem isn't the stolen accounts, it's your conversations.
Ten days after teaming up with France's Mistral, Saudi Arabia's state AI company has now built on a Chinese open model
Fei-Fei Li's World Labs has rolled generation, reconstruction, and robot simulation into a single model called Atlas
vLLM-Omni and FastVideo's FastH3 have pushed generation time below the video's own runtime
'Transfer station' proxies that defeat even selfie verification are undermining regional blocks and identity checks
Community tokenizer comparison shows 60/60 match with zero average error — but that doesn't confirm it's the same model
Enter a goal and the agent automatically produces ad and e-commerce assets — yet the obvious domain name is up for sale
Compiler included under Apache 2.0 — MiniMax already serving billions of tokens per minute
Open-weight MiniMax-Music3 generates a complete 32kHz stereo track in one pass from just two inputs: lyrics and a structured caption
Artificial Analysis measured 56 on its Intelligence Index and 1.7 minutes per task. It also led AA-AnalystAgent at 60%, ahead of Claude Opus 5.
v3.0.54 adds form-filling and file-upload automation plus long-task tracking
Instead of generating ten times to get one usable result, the tool lets users lock in scenes and timing before rendering
