Kimi K3.1 Pro ranks fourth on the Artificial Analysis Intelligence Index with a score of 55

Kimi K3.1 Pro ranks fourth on the Artificial Analysis Intelligence Index with a score of 55

Moonshot's Kimi K3.1 Pro has received a score of 55 on the Artificial Analysis Intelligence Index, placing it fourth overall. The model trails Claude Opus 5.5 (max with fallback) at 58, Claude Sonnet 5.5 (max with fallback) at 56, and Muse Wonder Full (max) at 56.

Kimi K3.1 Pro improves by 11 points over Kimi K3 (max), which scored 44. If Moonshot publishes the model's weights as expected, K3.1 Pro will become the highest-scoring open-weights model on the index. The gap between the leading open-weights and proprietary models would then narrow from 12 points to 3.

Agentic performance

The largest improvements appear in evaluations of agentic capabilities:

• GDPval-AA: Kimi K3.1 Pro ranks third with 1,846 Elo, an increase of 312 Elo over Kimi K3. It trails Claude Opus 5.5 at 1,866 Elo and Muse Wonder Full at 1,857 Elo.
• Terminal-Bench 4.0: K3.1 Pro achieves the highest score of any model evaluated.
• AutomationBench-AA: The model scores 73%, compared with 58% for Kimi K3.
• Terminal-Bench 4.0 result: K3.1 Pro scores 67%, up from 13% for Kimi K3 and ahead of Claude Sonnet 5.5 at 64%.

Other evaluation results

Improvements on several other evaluations are more moderate:

• Humanity's Last Exam rises from 47% to 52%.
• CritPt increases from 23% to 25%.
• SciCode improves from 59% to 61%.
• The AA-Omniscience Index rises from 20 to 30.
• The hallucination rate falls from 53% to 44%.
• Accuracy increases from 48% to 51%.

Cost and token usage

The cost per task is approximately the same as Kimi K3, at $1.98 per task. Kimi K3.1 Pro uses 55,000 output tokens per Intelligence Index task, 14% more than Kimi K3's 48,000 tokens. Its first-party token prices are 33% lower, offsetting the higher output-token usage.

Model details

• Context window: 1 million tokens
• Model size: 2.8 trillion total parameters, with 104 billion active parameters
• Input modalities: Text and image
• Weights and license: The weights were not published at launch. Moonshot is expected to release them soon after launch.
• Pricing: $2 per 1 million input tokens, $10 per 1 million output tokens, and $0.20 per 1 million cache-hit tokens. Each rate is 33% lower than Kimi K3's corresponding rate.

Summary of benchmark results

Kimi K3.1 Pro's overall score reflects particularly strong gains in agentic tasks, while its improvements on general knowledge, coding, and scientific evaluations are smaller. Its pending weight release will determine whether it becomes the highest-scoring open-weights model on the Artificial Analysis Intelligence Index.