DeepSeek-V4-Pro-0813
DeepSeek-V4-Pro-0813 is DeepSeek's flagship 1.6T-parameter MoE model released in August 2026. It is optimized for agentic coding, terminal automation and security evaluation, with leading results on Terminal-Bench 2.1 and CyberGym. It is available through multiple providers including DeepSeek, DeepInfra, Novita and Together.
Key Specifications
Parameters
1.6T
Context
1.0M
Release Date
August 13, 2026
Average Score
60.6%
Timeline
Key dates in the model's history
Announcement
August 13, 2026
Last Update
August 27, 2026
Today
September 8, 2026
Technical Specifications
Parameters
1.6T
Training Tokens
-
Knowledge Cutoff
-
Family
-
Capabilities
MultimodalZeroEval
Pricing & Availability
Input (per 1M tokens)
$1.32
Output (per 1M tokens)
$3.96
Max Input Tokens
1.0M
Max Output Tokens
393.2K
Supported Features
Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning
Benchmark Results
Model performance metrics across various tests and benchmarks
Other Tests
Specialized benchmarks
Terminal-Bench 2.1
DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95 • Self-reported
CyberGym
DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95 • Self-reported
Toolathlon
Verified split; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95 • Self-reported
DSBench-FullStack
Internal full-stack development test set; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95 • Self-reported
Agents' Last Exam
DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
AutomationBench
Public set; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
DeepSWE
DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
DSBench-Hard
Internal difficult coding-agent test set; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
Humanity's Last Exam
With tools; max reasoning effort. • Self-reported
Humanity's Last Exam (no tools, text-only)
HLE without tools; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
Humanity's Last Exam (with tools, text-only)
HLE with tools; DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
NL2Repo
DeepSeek Harness minimal mode; max reasoning effort; temperature 1.0; top_p 0.95. • Self-reported
License & Metadata
License
mit
Announcement Date
August 13, 2026
Last Updated
August 27, 2026
Compare DeepSeek-V4-Pro-0813
All comparisonsSimilar Models
All ModelsDeepSeek-R1-0528
DeepSeek
671.0B
Best score:0.8 (GPQA)
Released:May 2025
Price:$0.70/1M tokens
DeepSeek-V3.1
DeepSeek
671.0B
Best score:0.8 (GPQA)
Released:Jan 2025
Price:$0.27/1M tokens
DeepSeek-V4-Flash-0731
DeepSeek
304.0B
Released:Jul 2026
Price:$0.14/1M tokens
DeepSeek-V3.2 (Non-thinking)
DeepSeek
685.0B
Released:Nov 2025
Price:$0.28/1M tokens
DeepSeek-V4-Flash-Max
DeepSeek
284.0B
Best score:0.9 (GPQA)
Released:Apr 2026
Price:$0.14/1M tokens
DeepSeek-V4-Pro-Max
DeepSeek
1.6T
Best score:0.9 (GPQA)
Released:Apr 2026
Price:$1.74/1M tokens
DeepSeek-V2.5
DeepSeek
236.0B
Best score:0.9 (HumanEval)
Released:May 2024
Price:$2.00/1M tokens
DeepSeek-V3 0324
DeepSeek
671.0B
Best score:0.7 (GPQA)
Released:Mar 2025
Price:$0.28/1M tokens
Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.