Kimi K3

Available

Moonshot AIFrontierOpen weights

Moonshot's flagship open-weight agentic model and the largest open model released to date: a 2.8T-parameter MoE (896 experts, 16 active per token) using Kimi Delta Attention and Attention Residuals, with native multimodal input and a 1M-token context. Launched via API on Jul 16, 2026 at $3/$15 per Mtok (cached input $0.30); full open weights published to Hugging Face on Jul 26, 2026 — a day ahead of the announced Jul 27 target — under a Modified MIT license, making it freely downloadable and self-hostable.

Official page ↗Model card ↗

Specifications

License: Open weights · Modified MIT
Weights: Downloadable
Architecture: Mixture-of-Experts
Parameters: 2.8T · 50B active
Context window: 1.0M tokens
Max output: 32K tokens
Knowledge cutoff: —
Price (in / out, $/M): $3 / $15
Modalities: TextVisionCode

Benchmarks

SWE-bench Verifiedself-reported

76.8%

source ↗

GPQA Diamondself-reported

93.5%

source ↗

Vendor-reported figures are claims until independently verified. See methodology.