AI BriefingMiniMax Unveils M3
Key point
MiniMax has unveiled **MiniMax M3**, an open-weight model that combines coding, 1M context, and native multimodal capabilities.
Details
MiniMax M3 is the first open-weight model to combine coding and agentic capabilities, large-scale context extension, and native multimodality.
Key capabilities are as follows:
- Coding and agentic performance: It recorded strong benchmark performance, including SWE-Bench Pro 59.0%, Terminal Bench 2.1 66.0%, and MCP Atlas 74.2%.
- Context extension: Through MiniMax Sparse Attention technology, it supports up to 1M (1 million) tokens of context.
- Native multimodal: Multimodal capabilities were integrated from the design stage.
The model weights and technical report are expected to be released within about 10 days. To celebrate the current API launch, a 50% discount promotion is running for the first 7 days when using context of 512K or below.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Toss Bank Introduces 'Stylepack' Feedback Loop to Resolve Coding Agent Team Rule Violations