DeepSeek V3.1 vs DeepSeek V4-Pro

Best cell per row highlighted. Null means undisclosed — never counted as zero.

DeepSeek
DeepSeek V3.1GA
7.7
AI Panel
DeepSeek
DeepSeek V4-Propreview
8.5
AI Panel

Identity & lifecycle

ProviderDeepSeekDeepSeek
Family / tierDeepSeek V3 · LargeDeepSeek V4 · Pro
StatusGApreview
Released2025-08-202026-04-23
Knowledge cutoff2025-062026-02

Architecture & context

Context window128K1M
Max output tokens8K384K
Input modalitiestexttext
Reasoning modeoptionaloptional
Open weightsYesYes

Pricing (per Mtok)

Input$0.21$0.435
Output$0.79$0.87
Cached input$0.021$0.0036
Batch input
Free tierYesYes

Speed

Output speed (tok/s)
Time to first token (s)
Latency tiermediummedium

Trust & deployment

Trains on inputsYesYes
LicenseMITMIT
Clouds
Compliance

AI Panel scoring

Unified score7.78.5
Decision Maker7.58
Domain Strategist7.58.5
Finance Lead8.59.5
Domain Practitioner88.5
Power User7.57.5
Skeptic7.57.5
Value score8.89.7

Benchmarks

MMLU-Pro83.787.5
GPQA Diamond7490.1
AIME 202588
SWE-bench Verified6680.6
HumanEval76.8
LiveCodeBench8093.5
Aider Polyglot76.3
MRCR Long Context83.5
SimpleQA57.9
Humanity's Last Exam37.7
LMArena Coding Elo1287
Artificial Analysis Index52

Scores link to their sources. Missing cells mean the vendor hasn't disclosed a result — honesty over padding.