Moonshot's Kimi K3 To Be Largest Open Model With 2.8 Trillion Parameters, Has 1M Context Window

Moonshot AI announced its Kimi K3 model with 2.8 trillion total parameters, which the company said makes it the largest open-source model released to date. The model surpasses DeepSeek's V4-Pro at 1.6 trillion parameters and other recent Chinese lab releases including Xiaomi's 1.02 trillion parameter model and Z.AI's 744 billion parameter offering. Moonshot said this marks the ninth time in the past year a Kimi release has set a new record for open-source model scale, following Kimi K2 at 1 trillion parameters and subsequent K2.5, K2.6 and K2.7 Code updates. Kimi K3 ships with a 1 million token context window suited for software engineering, knowledge work and deep reasoning. The model uses a new architecture involving Attention Residuals, a technique Moonshot open-sourced earlier this year, with native visual understanding baked directly into the model. Moonshot said K3 delivers frontier-level performance ranking second only to Claude Fable 5 and GPT-5.6 Sol on its internal evaluation set. Pricing is set at $0.30 per million tokens for cached input, $3 per million tokens for uncached input and $15 per million tokens for output. Full model weights are expected to be released soon with a technical report to follow. The model is available on Kimi's website though benchmarks are currently awaited.