DeepSeek-V3.1

Name: DeepSeek-V3.1
Author: DeepSeek

DeepSeek

DeepSeek-V3.1 is a hybrid model supporting both thinking and non-thinking modes through different chat templates. Built on DeepSeek-V3.1-Base with two-phase long context extension (32K phase: 630B tokens, 128K phase: 209B tokens), it has 671B total parameters with 37B activated. Key improvements include smart tool calling, high thinking efficiency comparable to DeepSeek-R1-0528, and FP8 format.

Key Specifications

Parameters

671.0B

Context

163.8K

Release Date

January 9, 2025

Average Score

58.4%

Repository Model Weights

Timeline

Key dates in the model's history

Announcement / Last Update

January 9, 2025

Today

September 10, 2026

Technical Specifications

Parameters

671.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$0.27

Output (per 1M tokens)

$1.00

Max Input Tokens

163.8K

Max Output Tokens

163.8K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

Agent mode, Non-Thinking • Self-reported

66.0%

Reasoning

Logical reasoning and analysis

GPQA

Pass@1, mode without thinking • Self-reported

75.0%

Other Tests

Specialized benchmarks

SimpleQA

Mode thinking with • Self-reported

93.0%

MMLU-Redux

Mode without thinking • Self-reported

92.0%

MMLU-Pro

Mode without thinking • Self-reported

84.0%

Aider-Polyglot

Mode without thinking • Self-reported

68.0%

AIME 2024

Pass@1, Non-Thinking mode • Self-reported

66.3%

AIME 2025

Pass@1, Non-Thinking mode • Self-reported

49.8%

BrowseComp

Thinking mode with search agent • Self-reported

30.0%

BrowseComp-zh

Thinking mode with search agent • Self-reported

49.2%

CodeForces

Div1 Rating, Thinking mode • Self-reported

69.7%

HMMT 2025

Pass@1, Non-Thinking mode • Self-reported

33.5%

Humanity's Last Exam

Pass@1, Thinking mode, text-only subset • Self-reported

15.9%

LiveCodeBench

Pass@1, 2408-2505, Non-Thinking mode • Self-reported

56.4%

SWE-bench Multilingual

Agent mode, Non-Thinking • Self-reported

54.5%

Terminal-Bench

Terminus 1 framework, Non-Thinking • Self-reported

31.3%

License & Metadata

License

mit

Announcement Date

January 9, 2025

Last Updated

January 9, 2025

Similar Models

All Models

DeepSeek-V3 0324

DeepSeek

671.0B

Best score:0.7 (GPQA)

Released:Mar 2025

Price:$0.28/1M tokens

DeepSeek-V3

DeepSeek

671.0B

Best score:0.9 (MMLU)

Released:Dec 2024

Price:$0.27/1M tokens

DeepSeek R1 Zero

DeepSeek

671.0B

Best score:0.7 (GPQA)

Released:Jan 2025

DeepSeek-V3.2 (Thinking)

DeepSeek

685.0B

Best score:0.8 (GPQA)

Released:Nov 2025

Price:$0.28/1M tokens

DeepSeek-V3.2-Exp

DeepSeek

685.0B

Best score:0.8 (GPQA)

Released:Sep 2025

Price:$0.27/1M tokens

DeepSeek-R1

DeepSeek

671.0B

Best score:0.9 (MMLU)

Released:Jan 2025

Price:$3.00/1M tokens

DeepSeek-V4-Pro-0813

DeepSeek

1.6T

Released:Aug 2026

Price:$1.32/1M tokens

DeepSeek-R1-0528

DeepSeek

671.0B

Best score:0.8 (GPQA)

Released:May 2025

Price:$0.70/1M 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.