Moonshot Releases Kimi K3 Open Weights
Model tops Agent Arena and open-weight leaderboards in multiple categories.
Moonshot AI open-sourced the weights and technical report for Kimi K3. The model leads among open-weight entries in the Agent Arena by a wide margin over GLM-5.2 and claims the top position in Frontend Code and Text arenas as well. Multiple observers note the release includes native vision support along with production deployment guidance from the vLLM team. Early reactions highlight strong benchmark gains and unusually low inference cost relative to prior frontier-scale systems. The release is confirmed through official posts and independent leaderboard updates with no conflicting reports.
Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside…
Reactions from ranked influencers
25 postsMust read!
A definitive guide on K3 for inference infrastructure folks :D https://twitter.com/vllm_project/status/2081788542079123939
Been reading the weights of the most powerful open weights Ai model yet released. Page one starts strong: mostly negatives, a late run of zeros, and one unexpectedly large positive value.
Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside…
Big update: Among open-weight models, Kimi K3 (Max) is #1 in the Agent Arena with +9.75% net-improvement, surpassing GLM-5.2 (Max) at +7.12%, and landed the #1 spot across 5 signals (see below). Kimi K3 (Max) is also now #1 in open-weight in the Frontend Code (1682 pts) and…
Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside…
From the team:
So many exciting research now shared openly. Enjoy! https://twitter.com/kimi_moonshot/status/2081760186235289764
Kimi K3 weights have been released! Kimi K3 is now the leading open weights model at 57 in the Artificial Analysis Intelligence Index Moonshot has released the weights of their 2.6T parameter model under their 'Kimi K3 License' which we have labelled ‘Commercial Use…
Technical report: https://github.com/MoonshotAI/Kimi-K3/blob/main/k3_tech_report.pdf Weights on @huggingface: https://huggingface.co/moonshotai/Kimi-K3 Further benchmarks on Artificial Analysis:
