Frontier Log

DeepSeekModel

DeepSeek-V3.1

Made by
DeepSeek
Released
Announced in
DeepSeek-V3.1 Releaseapi-docs.deepseek.com
Event
DeepSeek-V3.1
API record
/api/v1/models/deepseek-v3-1

Benchmark results

BenchmarkScoreSource check
Humanity's Last ExamNo tools · reasoning mode; full or text-only set not stated · reported later, in DeepSeek-V3.1-Terminus15.9%
CorroboratedThe same number is printed in 2 other lab documents: 1 2
Humanity's Last ExamWith tools · search agent29.8%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
SWE-bench Verifiednon-thinking mode, as an agent66.0%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
GPQA Diamondthinking mode; 4,122 output tokens on average80.1%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
BrowseCompsearch agent30.0%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
AIME 2025No tools · thinking mode; 15,889 output tokens on average88.4%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Terminal-Bench 1.031.3%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
LiveCodeBenchthinking mode; 13,977 output tokens on average; problem window not stated74.8%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Sources

Reported by DeepSeek

Events

All models