Moonshot AI releases Kimi K3: 2.8T open-weight MoE model beating GPT-5.6 Sol and Claude on coding/browsing; weights coming July 27
Key Questions
What is Kimi K3 and its key specifications?
Moonshot AI released Kimi K3, a 2.8T parameter open-weight MoE model with 1M context and two variants (Max and Swarm). It targets coding and browsing tasks.
How does Kimi K3 compare to leading models?
It beats GPT-5.6 Sol and Claude on several coding and browsing benchmarks and claims to be the first open model in the 3T class. Real-world performance may vary from benchmarks.
When will the model weights be released?
Weights for Kimi K3 are scheduled for release on July 27, reinforcing the trend of Chinese open models closing the gap with proprietary US systems.
What variants are available?
Kimi K3 offers Max and Swarm variants built for different use cases within the 2.8T MoE architecture.
Why is this release significant?
It highlights accelerating progress by Chinese labs in open-weight frontier models, with benchmark wins that narrow the gap to top US proprietary systems like those from OpenAI and Anthropic.
Moonshot AI released Kimi K3, a massive 2.8T parameter open-weight MoE model with 1M context, two variants (Max and Swarm). It beats GPT-5.6 Sol and Claude on several coding and browsing benchmarks, claiming to be the first open model in the 3T class. Weights are scheduled for release on July 27. This reinforces the trend of Chinese open models closing the gap with proprietary US rivals. Real-world experience may lag, but benchmark wins are significant.