Moonshot AI has introduced its Kimi K3.1 Pro, a new flagship large language model available via the Kimi API. Coverage describes it as an upgraded, post-trained version of the earlier Kimi K3 released in July. Outlets also note a “Highspeed” variant marketed as faster, along with specific token pricing for the US dollar market. One outlet says the model is not yet offered as open weights, but that full weights are planned for release by Nov 4, 2026.
Reports provide performance and pricing details. MarkTechPost and Simon Willison both include benchmark and cost figures, including Terminal-Bench 4.0 results and per-million-token pricing for input and output. They also describe Kimi K3.1 Pro as a mixture-of-experts (MoE) model with a very large context window (1M tokens). Across sources, benchmark comparisons frame Kimi K3.1 Pro as broadly competitive with frontier models such as Claude Opus 5.5 and GPT-6 Astra, with some benchmarks where it is said to match or slightly outperform others.
