Artificials
Compare models
All makers

AI model maker

DeepSeek

38 models on Artificials, 24 with published test results and 34 with open weights you can download.

RSS feed
Best on the capability index155.3DeepSeek V4 Pro 0813, #29 of 274
Top score on1 testof the 25 its models are tested on
Newest releaseDeepSeek V4.1 FlashSep 10, 2026
Models3824 tested, 34 with open weights

Where DeepSeek stands

Its best model’s rank among all models tested on each skill area’s main test, the same tests model pages use. A longer bar means a better rank.

  • OverallDeepSeek V4 Pro 0813 on Epoch Capabilities Index
    Top 11%
  • CodingDeepSeek V4.1 Flash (max) on LiveBench coding
    Top 31%
  • MathDeepSeek V4 Pro 0813 (max) on Mock AIME 2024–2025
    Top 10%
  • KnowledgeDeepSeek V4 Pro 0813 (max) on SimpleQA Verified
    Top 28%
  • AgentsDeepSeek V4.1 Flash (max) on LiveBench agentic coding
    #1 of 63
  • LanguageDeepSeek V4 Pro 0813 on LiveBench language
    Top 39%
  • Data analysisDeepSeek V4 Flash Vision Exp on LiveBench data analysis
    Top 18%
  • ReasoningDeepSeek V4 Pro 0813 (max) on GPQA Diamond
    Top 12%

On the capability index

Its best 12 of 20 models on the Epoch Capabilities Index, each with its rank among all 274 models. The top score is 167.3, held by Claude Opus 5.5 from Anthropic.

Every test

Its best result on each published test, and that result’s rank among every model tested. Scores from different tests are never combined.

TestBest modelScoreRank
Epoch Capabilities IndexDeepSeek V4 Pro 0813155.3#29 of 274
LiveBench agentic codingDeepSeek V4.1 Flash (max)77.3%#1 of 63
Arena mathDeepSeek V4.1 Flash (max)1503#12 of 372
Arena instruction followingDeepSeek V4.1 Flash (max)1479#18 of 389
Arena codingDeepSeek V4.1 Flash (max)1528#22 of 384
MATH Level 5DeepSeek R1 052896.6%#7 of 97
Arena hard promptsDeepSeek V4.1 Flash (max)1500#30 of 389
Arena textDeepSeek V4.1 Flash (max)1474#32 of 389
Chess puzzlesDeepSeek V4 Pro 0813 (max)47.0%#13 of 141
Mock AIME 2024–2025DeepSeek V4 Pro 0813 (max)98.6%#19 of 198
Arena creative writingDeepSeek V4 Pro1446#39 of 387
LiveBench overallDeepSeek V4.1 Flash (max)81.1%#7 of 63
GPQA DiamondDeepSeek V4 Pro 0813 (max)91.7%#26 of 222
Arena WebDevDeepSeek V4.1 Flash (max)1620#18 of 123
LiveBench data analysisDeepSeek V4 Flash Vision Exp79.5%#11 of 63
SWE-bench VerifiedDeepSeek V4 Pro (max)77.6%#6 of 32
LiveBench mathematicsDeepSeek V4 Pro 081395.1%#13 of 63
SimpleQA VerifiedDeepSeek V4 Pro 0813 (max)52.9%#21 of 77
SWE-bench Verified, bash onlyDeepSeek V3.2 (high)70.0%#12 of 42
LiveBench codingDeepSeek V4.1 Flash (max)80.0%#19 of 63
LiveBench languageDeepSeek V4 Pro 081382.1%#24 of 63
FrontierMath Tiers 1–3DeepSeek V4 Pro 0813 (max)64.6%#31 of 81
LiveBench instructionsDeepSeek V4 Flash Vision Exp71.0%#25 of 63
LiveBench reasoningDeepSeek V4.1 Flash (max)86.7%#28 of 63
FrontierMath Tier 4DeepSeek V4 Pro 0813 (max)26.8%#32 of 63

Every model

Newest first. The price is per 1M tokens, blended three input tokens to one output token: DeepSeek’s own listing, or else the middle price across hosts.

ModelReleasedContextPriceCapability index
DeepSeek V4.1 FlashSep 10, 20261M tokens$0.385154.9
DeepSeek V4 FlashSep 10, 20261M tokens$0.263146.1
DeepSeek Flash LatestSep 10, 20261M tokens$0.334Not on the index
DeepSeek V4 Flash Vision ExpSep 10, 20261M tokens$0.263Not on the index
DeepSeek V4.1 FlashSep 10, 20261M tokens$0.263Not on the index
deepseek-ai/DeepSeek-V4.1-Flash-FastSep 10, 20261M tokens$0.525Not on the index
DeepSeek V4 Pro 0813Aug 12, 20261M tokens$1.98155.3
DeepSeek V4 ProAug 12, 20261M tokens$0.99149.1
DeepSeek V4 Flash 0731Jul 31, 20261M tokens$0.175154.5
DeepSeek V4 Flash 0731 FastJul 31, 20261M tokens$0.438Not on the index
DeepSeek V4 Flash LatestJul 31, 20261M tokens$0.175Not on the index
DeepSeek Pro LatestJul 6, 20261M tokens$1.23Not on the index
DeepSeek V4 Flash FreeApr 24, 20261M tokensNo paid listingNot on the index
DeepSeek V4 Flash 0423Apr 23, 20261M tokens$0.174Not on the index
DeepSeek V4 Pro 0423Apr 23, 20261M tokens$1.98Not on the index

Models, prices, context windows and release dates from models.dev (MIT). Test results from LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Arena (CC BY 4.0) and SWE-bench (CC BY-NC 4.0). A model counts here when DeepSeek sells it, several hosts list it or a test covers it. Each figure is the source’s own; nothing is combined into a new score.