DeepSeek

DeepSeek-V3.1

Zero-eval
#1BrowseComp-zh
#2SWE-bench Multilingual
#3SimpleQA

by DeepSeek

+
+
+
+
About

DeepSeek-V3.1 was developed as an incremental advancement over DeepSeek-V3, designed to refine the mixture-of-experts architecture with improved training techniques. Built to enhance quality and efficiency while maintaining the open-source philosophy, it represents continued iteration on DeepSeek's flagship model line.

+
+
+
+
Pricing Range
Input (per 1M)$0.27 -$0.27
Output (per 1M)$1.00 -$1.00
Providers2
+
+
+
+
Timeline
AnnouncedJan 10, 2025
ReleasedJan 10, 2025
+
+
+
+
License & Family
License
MIT
Base ModelDeepSeek-V3
Performance Overview
Performance metrics and category breakdown

Overall Performance

16 benchmarks
Average Score
58.4%
Best Score
93.4%
High Performers (80%+)
3

Performance Metrics

Max Context Window
327.7K
+
+
+
+
All Benchmark Results for DeepSeek-V3.1
Complete list of benchmark scores with detailed information
SimpleQA
text
0.93
93.4%
Self-reported
MMLU-Redux
text
0.92
91.8%
Self-reported
MMLU-Pro
text
0.84
83.7%
Self-reported
GPQA
text
0.75
74.9%
Self-reported
CodeForces
text
0.70
69.7%
Self-reported
Aider-Polyglot
text
0.68
68.4%
Self-reported
AIME 2024
text
0.66
66.3%
Self-reported
SWE-Bench Verified
text
0.66
66.0%
Self-reported
LiveCodeBench
text
0.56
56.4%
Self-reported
SWE-bench Multilingual
text
0.55
54.5%
Self-reported
Showing 1 to 10 of 16 benchmarks
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+