Kimi K3
Kimi · 2 thinking configurations
Intelligence43.6
Input / 1M$3.00
Output / 1M$15.00
Context1M
Tariff: Kimi · Source: artificial-analysis · USD per million tokens
Thinking configuration
What does more thinking buy you?
Compare this model’s measured thinking configurations.
Scroll the chart ↔ · tap a point for details
Efficient frontier: best measured score at each cost1 measured configurations · 1 without both values
Selected: thinking max · Click a point to switch configuration
Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...
Intelligence IndexArtificial Analysis composite index, not a percentage. Different versions of the index are not directly comparable.
43.6 pointsDeepSWE · pass@1Software-engineering tasks on mini-swe-agent v1.1. Pass@1 across evaluated runs, with measured cost per task. The 95% intervals may overlap.
68.5 % ±4.5GPQA DiamondGraduate-level scientific questions. Measures specialist reasoning rather than web-search quality.
93.5 %Humanity’s Last ExamDifficult questions across many disciplines, compared under the same evaluation setup.
46.9 %SciCodeCode generation for scientific research problems.
59.5 %Terminal-Bench 2.1Agent tasks in a terminal environment. Results also depend on the evaluation harness.
85.0 %MMMU-ProMultimodal understanding of academic problems that include images.
80.5 %Omniscience accuracyFactual accuracy in Artificial Analysis's Omniscience evaluation.
47.6 %Specifications from the catalog
Inputtext, image, video
Outputtext
Tool callingYes
Structured outputYes
Speed36.3 tok/s
First token3.75 s
Maximum output943.7K
Catalog listing date2026-07-16
DeepSWE v1.1 · 113 tasks · mini-swe-agent · cost $4.65 / task · 98 mean steps · 81.5K mean output tokens. Published by the source 9/22/2026.