Apex
We show this table for reference; we do not rank on it.
A high-difficulty mathematical reasoning benchmark reported in DeepSeek-V4 model evaluations.
Benchmark score on Apex — October 29, 2026
We compile the Apex rows from provider self-reports. Hy4 preview leads the table at 74.2%, followed by DeepSeek V4.1 Flash 1017 (65.6%) and A.X K2 (45.8%). We do not use these results to rank models overall.
Hy4 preview
Tencent
DeepSeek V4.1 Flash 1017
DeepSeek
A.X K2
SK Telecom
12 modelsMathematicsCurrentDisplay onlyUpdated October 29, 2026
Benchmark score table (12 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Hy4 previewTencent | 74.2% | Not reported | Open |
| 2 | DeepSeek V4.1 Flash 1017DeepSeek | 65.6% | Not reported | Open |
| 3 | A.X K2SK Telecom | 45.8% | Not reported | Open |
| 4 | Qwen3.7 MaxAlibaba | 44.5% | Not reported | Closed |
| 5 | DeepSeek V4 Pro 0813DeepSeek | 38.3% | Not reported | Open |
| 6 | DeepSeek V4 Flash 0731DeepSeek | 33.0% | Not reported | Open |
| 7 | ZAYA1-8BZyphra | 32.2% | Not reported | Open |
| 8 | DeepSeek V4 Pro (High)DeepSeek | 27.4% | Not reported | Open |
| 9 | Qwen3.7 PlusAlibaba | 22.7% | Not reported | Closed |
| 10 | DeepSeek V4 Flash (High)DeepSeek | 19.1% | Not reported | Open |
| 11 | DeepSeek V4 FlashDeepSeek | 1.0% | Not reported | Open |
| 12 | DeepSeek V4 ProDeepSeek | 0.4% | Not reported | Open |
Among the reported Apex rows, Hy4 preview is first at 74.2%. The third row is 28.4 points behind. The broader top-10 range is 55.1 points, so the table still separates the published systems.
12 models have been evaluated on Apex. The benchmark falls in the Mathematics category. Apex is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Apex
Year
2026
Tasks
Advanced mathematical reasoning
Format
Pass@1 math benchmark
Difficulty
Frontier math reasoning
BenchLM stores Apex as a display-only provider-table reference for exact first-party model comparisons.
Freshness and provenance
Version
Apex 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 Apex measure?
A high-difficulty mathematical reasoning benchmark reported in DeepSeek-V4 model evaluations.
Which model scores highest on Apex?
Hy4 preview by Tencent currently leads with a score of 74.2% on Apex.
How many models are evaluated on Apex?
12 AI models have published results on Apex 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.