Moonshot AI has released Kimi K3, an open-weights model with approximately 2.8 trillion parameters and a 1 million token context window. The model achieved strong performance on benchmarks like Terminal-Bench 2.0, FrontierSWE, and GPQA Diamond. Notably, Kimi K3 has been ranked above Claude 3 Opus in Arena blind tests for text tasks and leads in front-end coding capabilities, surpassing all US frontier models. The release also features competitive pricing at $3/$15 per million tokens, positioning it as a frontier-level model despite its open-weight status. AI
IMPACT Sets a new bar for open-weight models, potentially accelerating adoption of large-context windows and challenging existing frontier models on price and performance.
RANK_REASON Frontier-lab model release with open weights and performance metrics. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — sigmoid.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
