Moonshot prices Kimi K3.1 Pro at half Claude Opus 5.5’s rate
Moonshot AI released Kimi K3.1 Pro, a post-trained update to its Kimi K3 model, at €1.85 ($2) per million input tokens and €9.25 ($10) per million output tokens. That is half Claude Opus 5.5’s list price, while matching Claude Sonnet 5.5’s per-token rates; Moonshot says the full model weights will be released by 4 November 2026.
Bottom line — The price gap is clear; independent coding tests disagree on whether Kimi’s extra reasoning makes it cheaper per task.
Go deeper 8
-
Moonshot reports 65.7% on Terminal-Bench 4.0, against 66.4% for Opus 5.5; OrcaRouter notes the scores came from different settings and vendor-controlled setups.
-
Bito’s coding tests scored Kimi 72%, Sonnet 75% and Opus 87%; a long Kimi session cost €3.95 ($4.28), versus €1.18 ($1.28) for Sonnet.
-
Atomic Agent found Kimi’s cost on one coding task ranged from €0.15–€0.19 ($0.16–$0.21) at low reasoning effort to €0.32–€0.43 ($0.35–$0.46) at the default maximum. Its three-run test is limited to one small Python task.
-
Kimi K3.1 Pro has a one-million-token context window and charges a flat rate across it, according to Moonshot’s pricing details reported by DEV Community.
-
Moonshot’s Highspeed version costs €3.70 ($4) per million input tokens and €18.50 ($20) per million output tokens; the company says it runs two to three times faster.
-
Artificial Analysis ranked Kimi fourth of 215 models on its Intelligence Index, with a score of 55. The Latent Daily says the tracker found it slower than average and more verbose than the median model.
-
Moonshot says it removed cyber-related data before release; its internal test score fell from 90.5% to 81.0%. The Latent Daily says these figures have not been independently replicated.
-
Moonshot says the full weights are due by 4 November 2026; they were not available on Hugging Face on launch day, Atomic Agent reports.