Compare · ModelsLive · 2 picked · head to head
Kimi K2.5 vs Kimi K3
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Kimi K3 wins on 10/11 benchmarks
Kimi K3 wins 10 of 11 shared benchmarks. Leads in agentic · reasoning · knowledge.
Category leads
speed·Kimi K2.5agentic·Kimi K3reasoning·Kimi K3knowledge·Kimi K3math·Kimi K3coding·Kimi K3
Hype vs Reality
Attention vs performance
Kimi K2.5
#122 by perf·#17 by attention
Kimi K3
#51 by perf·#17 by attention
Best value
Kimi K2.5
4.3x better value than Kimi K3
Kimi K2.5
37.0 pts/$
$1.35/M
Kimi K3
8.7 pts/$
$6.86/M
Vendor risk
Who is behind the model
Moonshot AI
$18.0B·Tier 1
Moonshot AI
$18.0B·Tier 1
Head to head
11 benchmarks · 2 models
Kimi K2.5Kimi K3
Artificial Analysis · Quality Index
Kimi K2.5 leads by +3.2
Kimi K2.5
46.8
Kimi K3
43.6
APEX-Agents
Kimi K3 leads by +36.2
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Kimi K2.5
14.4
Kimi K3
50.6
ARC-AGI
Kimi K3 leads by +29.2
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Kimi K2.5
65.3
Kimi K3
94.5
ARC-AGI-2
Kimi K3 leads by +48.6
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Kimi K2.5
11.8
Kimi K3
60.4
Chess Puzzles
Kimi K3 leads by +28.4
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Kimi K2.5
7.4
Kimi K3
35.8
GPQA diamond
Kimi K3 leads by +7.3
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Kimi K2.5
83.5
Kimi K3
90.8
OTIS Mock AIME 2024-2025
Kimi K3 leads by +5.0
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Kimi K2.5
92.2
Kimi K3
97.2
PostTrainBench
Kimi K3 leads by +21.7
Kimi K2.5
10.3
Kimi K3
32.0
SimpleBench
Kimi K3 leads by +16.7
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Kimi K2.5
36.2
Kimi K3
52.8
SimpleQA Verified
Kimi K3 leads by +16.3
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Kimi K2.5
34.3
Kimi K3
50.6
WeirdML
Kimi K3 leads by +37.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Kimi K2.5
45.6
Kimi K3
82.6
Full benchmark table
| Benchmark | Kimi K2.5 | Kimi K3 |
|---|---|---|
Artificial Analysis · Quality Index | 46.8 | 43.6 |
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 14.4 | 50.6 |
ARC-AGI ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization. | 65.3 | 94.5 |
ARC-AGI-2 ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data. | 11.8 | 60.4 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 7.4 | 35.8 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 83.5 | 90.8 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 92.2 | 97.2 |
PostTrainBench | 10.3 | 32.0 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 36.2 | 52.8 |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 34.3 | 50.6 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 45.6 | 82.6 |
Pricing · per 1M tokens · projected $/mo at 10M tokens