Kimi is releasing K3, a new open-weight multimodal model featuring 2.8 trillion parameters and a context window of one million tokens. In internal benchmarks, K3 demonstrates performance competitive with models like Claude Fable 5 and GPT 5.6 Sol, surpassing others such as Opus 4.8 and GLM 5.2. This release also indicates a potential shift towards higher costs for Chinese AI models, with full model weights expected by July 27. AI
IMPACT Sets a new benchmark for open-weight models with large context windows, potentially driving competition and cost increases in the Chinese AI market.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
