Hy3-preview-Base is the stronger all-rounder.
Scores updated · 20 tests both models report · How we compare
Where each one wins
Tests won in each of the one area 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 rests on a single test.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Hy3-preview-Base pulls ahead
- Recent programming contest problemsLiveCodeBench v6+5.6points ahead
Where DeepSeek-V3-Base pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results19 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.
- MATHHy3-preview-Base by 16.959.476.3+16.9
- GSM8KHy3-preview-Base by 7.288.295.4+7.2
- CMathHy3-preview-Base by 5.785.591.2+5.7
- SuperGPQAHy3-preview-Base by 5.446.251.6+5.4
- CRUXEval-I (input prediction)Hy3-preview-Base by 3.467.871.2+3.4
- MBPP+ (EvalPlus-augmented)Hy3-preview-Base by 3.275.578.7+3.2
- CRUXEval-O (output prediction)DeepSeek-V3-Base by 2.67168.4+2.6
- MMLU-ProHy3-preview-Base by 1.86465.8+1.8
- CMMLUHy3-preview-Base by 1.787.989.6+1.7
- ARC-ChallengeHy3-preview-Base by 1.394.796+1.3
- Chinese SimpleQA (C-SimpleQA)Hy3-preview-Base by 168.769.7+1
- DROPDeepSeek-V3-Base by 186.585.5+1
- Includetie77.978.6tie
- MMMLUtie79.580.2tie
- C-Evaltie90.389.8tie
- MMLUtie87.787.4tie
- SimpleQAtie26.126.5tie
- PIQAtie84.284.4tie
- MMLU-Reduxtie86.886.9tie
Questions people ask
Which is better, DeepSeek-V3-Base or Hy3-preview-Base?
Hy3-preview-Base wins the one area where both have results: coding. DeepSeek-V3-Base wins none.
Which is better for coding?
Hy3-preview-Base. It wins the one coding test both models report.
How do you compare the two?
We use the 20 benchmark tests both models have published scores on. The verdict counts the 1 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.