Benchmark

DeepSeek — reported benchmarks

DeepSeek · source page ↗ · last checked Oct 4, 2026, 12:04 PM

Reported benchmarks · DeepSeek-V4.1-Flash

captured Oct 4, 2026, 12:04 PM
BenchmarkScore
GPQA Diamond90.9% · Pass@1
Terminal-Bench 2.190.6%
Codeforces3471 Rating
DeepSWE v1.174.2%
NL2Repo-Bench65.4%
CyberGym88.1%
HLE36.8% · without tools; 39.1% with tools
MathArena Apex65.6%
Terminal-Bench 3.030.0%
Terminal-Bench 4.031.2%
Agents' Last Exam31.8%

Vendor-reported via automated web search — not independently verified. See the cited matrix on /models.

In the news · DeepSeek

Importance-filtered press coverage (Google News) mentioning DeepSeek. Headlines link to the original; verify before acting.

Change history