CMath
We show this table for reference; we do not rank on it.
A Chinese mathematical reasoning benchmark reported in DeepSeek-V4 base-model evaluations.
Benchmark score on CMath — October 29, 2026
We compile the CMath rows from provider self-reports. DeepSeek V4 Flash Base leads the table at 93.6%, followed by DeepSeek V4 Pro Base (90.9%). We do not use these results to rank models overall.
DeepSeek V4 Flash Base
DeepSeek
DeepSeek V4 Pro Base
DeepSeek
2 modelsMathematicsCurrentDisplay onlyUpdated October 29, 2026
Benchmark score table (2 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | DeepSeek V4 Flash BaseDeepSeek | 93.6% | Not reported | Open |
| 2 | DeepSeek V4 Pro BaseDeepSeek | 90.9% | Not reported | Open |
About CMath
Year
2026
Tasks
Chinese math problems
Format
Exact match
Difficulty
Math reasoning
BenchLM stores CMath as a display-only provider-table row when exact values are published in DeepSeek-V4 evaluations.
Freshness and provenance
Version
CMath 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does CMath measure?
A Chinese mathematical reasoning benchmark reported in DeepSeek-V4 base-model evaluations.
Which model scores highest on CMath?
DeepSeek V4 Flash Base by DeepSeek currently leads with a score of 93.6% on CMath.
How many models are evaluated on CMath?
2 AI models have published results on CMath in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.