DeepSeekopen weightsCurrentReasoning
DeepSeek V4 Pro
Scores are the 0813 checkpoint (GA Aug 13, 2026) at max reasoning unless noted.
Apr 24, 2026
$0.435 / $0.87
$0.544
1M
Benchmark scores
| Benchmark | Score | Δ vs prior | Rank (current) | Configuration | Source |
|---|---|---|---|---|---|
Terminal-Bench 2.1 | 54.7% | #17 of 20 | 0813, max reasoning · leaderboard | benchlm.ai ↗ | |
SWE-bench Verified | 80.6% | #2 of 8 | 0813 · leaderboard | benchlm.ai ↗ | |
GPQA Diamond | 90.1% | #7 of 10 | 0813 · leaderboard | benchlm.ai ↗ | |
Humanity's Last Exam | 42.7% | #9 of 14 | 0813 · leaderboard | benchlm.ai ↗ | |
ARC-AGI-2 | 61.3% | #7 of 9 | 0813 · leaderboard | benchlm.ai ↗ | |
τ²-bench (Telecom) | 96.2% | #1 of 9 | 0813 · leaderboard | benchlm.ai ↗ | |
MMLU-Pro | 87.5% | #1 of 3 | 0813 · leaderboard | benchlm.ai ↗ | |
Artificial Analysis Intelligence Index | 44.3 | #4 of 10 | max · leaderboard | benchlm.ai ↗ |