Kimi K3 (max) shows strong intelligence with high costs

Kimi K3 (max), an open weights reasoning model from Kimi released on July 16, 2026, scores 57 on the Artificial Analysis Intelligence Index. The result places it well above comparable open weight models of similar size, where the median is 25. It supports text and image input, produces text output, and offers a 1M tokens context window.

The model is comparatively expensive for its class, priced at $3.00 per 1M input tokens and $15.00 per 1M output tokens, with cache hits priced at $0.30 per 1M tokens. Running the Intelligence Index evaluation cost $2437.41, and a blended rate using a 7:2:1 cache hit/input/output ratio is listed at $2.31 per 1M tokens.

Performance is mixed beyond intelligence. Kimi K3 (max) generates output at 34.9 tokens per second and has a time to first token (TTFT) of 3.94s, both slower than the reported medians of 61.6 t/s and 1.87s. It generated 130M output tokens during the Intelligence Index evaluation, reflecting high verbosity. The model has 2.8 trillion parameters, with 104 billion active during inference, and is released under the Kimi K3 License license, with commercial use requiring a separate agreement.