← All models
Compare

DeepSeek-V3.2 vs Kimi K2.5

Add, remove, or swap models to compare them side by side.

DeepSeek-V3.2Kimi K2.5
DeepSeek-V3.2DeepSeek · 1 in the newsKimi K2.5Moonshot · 10 in the news
Scores
Intelligence (ECI)146148
Coding2746
Math3653
Reasoning & Knowledge2637
Agentic & Tools2939
Specifications
DeveloperDeepSeekMoonshot
FamilyDeepSeekKimi
ReleasedDec 1, 2025Feb 2, 2026
Parameters1T
AvailabilityOpen weights (unrestricted)Open weights (unrestricted)
Context window131K262K
Price — $/M input$0.23$0.38
Price — $/M output$0.34$2.02
Inputstexttext, image
Outputstexttext
Benchmarks
AIME 2024/202588%92%
APEX7%14%
ARC-AGI57%65%
ARC-AGI-24%12%
FrontierMath22%28%
FrontierMath Tier 42%4%
GPQA Diamond83%88%
SimpleBench53%47%
SimpleQA Verified28%34%
Terminal-Bench40%43%
WebDev Arena12861431
WeirdML47%46%
Humanity's Last Exam24%
SWE-bench Verified74%

Use-case scores are 0–100 percentile composites across each area’s benchmarks, ranked against every model from the past year. Highlighted cells lead each row. Open a model for the full picture.

Benchmarks & model data from Epoch AI (CC BY); pricing & specs from OpenRouter. ECI = Epoch Capabilities Index.