MiniMax, an AI firm from Shanghai, has launched an open-source reasoning model, MiniMax-M1, which is positioned to compete with top firms such as DeepSeek, OpenAI, and Google. Launched under an Apache license, M1 boasts superior capabilities in complex scenarios and contains a context window handling up to one million tokens, significantly outdoing competitors. The model has shown competitive benchmarks against other major models, although results should be verified independently. This release marks MiniMax's effort to dominate the AI space by providing industry-leading cost-effectiveness and performance.
"In complex, productivity-oriented scenarios, M1's capabilities are top-tier among open-source models, surpassing domestic closed-source models and approaching the leading overseas models, all while offering the industry's best cost-effectiveness."
"MiniMax makes clear that it's trying to supplant DeepSeek as the leading industry disruptor by noting that its context window is one million tokens, which rivals Google Gemini 2.5 Pro and is eight times the capacity of DeepSeek R1."
Collection
[
|
...
]