Kimi K2 Base is the stronger all-rounder.
Scores updated · 50 tests both models report · How we compare
Where each one wins
Tests won in each of the two areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
Coding and reasoning rest on a single test each.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Kimi K2 Base pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+16.3points ahead
- Recent programming contest problemsLiveCodeBench v6+1.4points ahead
Where DeepSeek-V3.2-Exp-Base pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results48 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- GSM-Infinite Hard (32K)DeepSeek-V3.2-Exp-Base by 19.145.226.1+19.1
- HumanEval+Kimi K2 Base by 17.167.784.8+17.1
- GSM-Infinite Hard (Tools-allowed)DeepSeek-V3.2-Exp-Base by 16.925.78.8+16.9
- GSM-Infinite Hard (128K)DeepSeek-V3.2-Exp-Base by 16.925.78.8+16.9
- GSM-Infinite Hard (64K)DeepSeek-V3.2-Exp-Base by 16.632.616+16.6
- HumanEvalKimi K2 Base by 16.361.978.2+16.3
- GSM-Infinite Hard (16K)DeepSeek-V3.2-Exp-Base by 15.850.434.6+15.8
- MultiPL-E HumanEvalKimi K2 Base by 14.845.760.5+14.8
- NIAH-Multi (32K)Kimi K2 Base by 14.285.699.8+14.2
- NIAH-Multi (64K)Kimi K2 Base by 14.185.9100+14.1
- MBPPKimi K2 Base by 13.475.689+13.4
- MBPP-SanitizedKimi K2 Base by 13.458.772.1+13.4
- MATHKimi K2 Base by 10.160.170.2+10.1
- SimpleQAKimi K2 Base by 8.32735.3+8.3
- MultiPL-E MBPPKimi K2 Base by 8.250.658.8+8.2
- GSM8KKimi K2 Base by 7.784.492.1+7.7
- aime_2024_2025_combinedKimi K2 Base by 6.824.831.6+6.8
- Chinese SimpleQA (C-SimpleQA)Kimi K2 Base by 6.66874.6+6.6
- MMLU-ProKimi K2 Base by 5.963.369.2+5.9
- GPQA (unspecified)Kimi K2 Base by 5.837.343.1+5.8
- CRUXEval-O (output prediction)DeepSeek-V3.2-Exp-Base by 5.374.969.6+5.3
- HellaSwagKimi K2 Base by 5.289.494.6+5.2
- NIAH-Multi (Tools-allowed)Kimi K2 Base by 5.294.399.5+5.2
- NIAH-Multi (128K)Kimi K2 Base by 5.294.399.5+5.2
- CRUXEval-I (input prediction)Kimi K2 Base by 4.163.968+4.1
- MBPP+Kimi K2 Base by 469.873.8+4
- MBPP+ (EvalPlus-augmented)Kimi K2 Base by 469.873.8+4
- RULER 128KDeepSeek-V3.2-Exp-Base by 3.391.988.6+3.3
- DROPDeepSeek-V3.2-Exp-Base by 386.683.6+3
- MGSMKimi K2 Base by 2.982.385.2+2.9
- OpenBookQAKimi K2 Base by 2.648.250.8+2.6
- AGIEval-EnKimi K2 Base by 2.570.172.5+2.5
- CMMLUKimi K2 Base by 288.990.9+2
- WinoGrandeKimi K2 Base by 1.983.485.3+1.9
- C-EvalKimi K2 Base by 1.59192.5+1.5
- IncludeDeepSeek-V3.2-Exp-Base by 1.577.275.7+1.5
- GlobalMMLUDeepSeek-V3.2-Exp-Base by 1.38280.7+1.3
- BigCodeBenchDeepSeek-V3.2-Exp-Base by 1.262.961.7+1.2
- RACEDeepSeek-V3.2-Exp-Base by 1.293.292+1.2
- TriviaQAKimi K2 Base by 1.283.985.1+1.2
- SuperGPQAKimi K2 Base by 1.143.644.7+1.1
- ARC-ChallengeKimi K2 Base by 195.296.2+1
- RULER 64Ktie93.393.8tie
- MMLU-Reduxtie90.490.2tie
- PIQAtie85.184.9tie
- BBHtie88.788.7tie
- Global-MMLU-Litetie85.685.6tie
- MMLUtie87.887.8tie
Questions people ask
Which is better, DeepSeek-V3.2-Exp-Base or Kimi K2 Base?
Kimi K2 Base wins all two areas where both have results: coding and reasoning. DeepSeek-V3.2-Exp-Base wins none.
Which is better for coding?
Kimi K2 Base. It wins the one coding test both models report.
How do you compare the two?
We use the 50 benchmark tests both models have published scores on. The verdict counts the 2 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 48 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.