Model release
Moonshot AI launches Kimi K3, a 2.8T-parameter open-weight model
Moonshot AI unveiled Kimi K3, a Mixture-of-Experts model with roughly 2.8 trillion total parameters, a 1-million-token context window, native vision, and always-on reasoning, aimed at long-horizon coding and agent workloads. Moonshot claims it performs competitively with Claude Fable 5 and beats Opus 4.8 and GPT-5.6 Sol on its benchmarks.
- Why it matters
- An open-weight model claiming near-frontier performance changes the cost calculus for teams currently paying premium API rates for agentic coding workloads — it's a real alternative to evaluate, not just a benchmark headline.
- Who should care
- Teams building AI coding agents or long-context reasoning workflows
- What you can do
- If you're evaluating models for agentic coding, benchmark Kimi K3 against your current model on your own workload before switching — vendor benchmark claims should be verified independently.
