Kimi K3.1 Pro Hits No. 4, Just Behind Claude Opus 5.5
Moonshot AI's model scores 55 points on the Artificial Analysis Intelligence Index, and its weights are slated for release by November 4, 2026.
In short
Moonshot AI's Kimi K3.1 Pro ranks fourth on the Artificial Analysis Intelligence Index with 55 points, three points behind Anthropic's Claude Opus 5.5.
At a glance
- 55 points on the Artificial Analysis Intelligence Index, rank 4 — an 11-point jump over K3.
- Only Claude Opus 5.5 (58), Claude Sonnet 5.5 (56) and Muse Wonder Full (56) score higher.
- 2.8 trillion total parameters, 104 billion of them active, 1 million token context window.
- Prices cut by 33 percent: $2.00 input and $10.00 output per million tokens.
- Weights slated for November 4, 2026; the open-to-proprietary gap would fall from 12 to 3 points.
Moonshot AI's Kimi K3.1 Pro scores 55 points on the Artificial Analysis Intelligence Index, which puts it in fourth place. That is 11 points above its predecessor, K3. Only three models from Anthropic and Meta rank higher: Claude Opus 5.5 at 58 points, Claude Sonnet 5.5 at 56 and Muse Wonder Full at 56.
A split scorecard
The Chinese model does not trail everywhere. It posts 51.7 on SWE Marathon v1.1 against 46.3 for Claude Opus 5.5, and 94.2 on GPQA-Diamond against 93.9. It also leads OfficeQA Pro v2 with 64.4 percent. Anthropic keeps the edge on Terminal-Bench 4.0, 66.4 percent to 65.7 percent, and on GDPval-AA v2.1, 1,866 Elo to 1,846.
Size, context and what it costs
The model carries 2.8 trillion parameters in total, 104 billion of them active per token, handles text and images, and accepts a context window of 1 million tokens. Moonshot cut API prices by 33 percent to $2.00 per million input tokens, $10.00 per million output tokens and $0.20 for cached input. Per task in the Intelligence Index that works out to $1.98, essentially level with K3 at $2.00, because the new model emits 14 percent more output tokens.
Fewer invented claims, by the vendor's own count
Moonshot reports that claim-level hallucinations on an internal benchmark fell from 23.4 percent with K3 to 8.3 percent. That figure comes from the company itself and has not been independently reproduced.
The release that would matter
Moonshot plans to publish the model weights by November 4, 2026. If that happens, K3.1 Pro becomes the highest-scoring open-weights model in the index, and the gap between open and proprietary frontier models narrows from 12 to 3 points. So far it is an announcement: the license terms, the exact release date and the scores of the weights actually published remain unconfirmed.
FAQ
How many points does Kimi K3.1 Pro score on the Artificial Analysis Intelligence Index?
55 points, good for fourth place. That is 11 points more than its predecessor K3 and three points less than Claude Opus 5.5 at 58 points.
When will the Kimi K3.1 Pro weights be released?
Moonshot AI is targeting November 4, 2026. Until the files are out this remains an announcement, and the license terms are not known.
How much does Kimi K3.1 Pro cost through the API?
$2.00 per million input tokens, $10.00 per million output tokens and $0.20 for cached input, roughly 33 percent below the previous prices.
