ResearchLLM Verifications Leaderboard

Data as of · 78 models, 80 questions · reviewed by Suhas Bhairav · methodology

DeepSeekOpen weightsReleased 12 Aug 2026

DeepSeek V4 Pro 0813 benchmarks, speed and API pricing

Reviewed by Suhas Bhairav · Tested

Compare

Available through OpenRouter on

  • Baidu$0.13992 in · $0.41976 out
  • Sail Research$0.45 in · $0.8 out
  • Alibaba$0.5808 in · $1.7424 out
  • StreamLake$0.66 in · $1.98 out
  • DeepSeek$0.66 in · $1.98 out
  • DeepInfra$1.3 in · $2.6 out
  • Phala$0.957 in · $2.8776 out
  • Ionstream$0.37 in · $2.93 out
  • Novita$0.99 in · $2.97 out
  • GMICloud$1.056 in · $3.168 out
  • DekaLLM$1.08 in · $3.8 out
  • CoreWeave$1.31 in · $3.96 out
  • BaseTen$1.32 in · $3.96 out
  • AtlasCloud$1.32 in · $3.96 out
  • Nebius$1.32 in · $3.96 out
  • Parasail$1.32 in · $3.96 out
  • Together$1.32 in · $3.96 out
  • DigitalOcean$1.32 in · $3.96 out
  • SiliconFlow$1.32 in · $3.96 out
  • Cloudflare$1.32 in · $3.96 out
  • Wafer$0.285 in · $4.3 out
  • Venice$1.65 in · $4.95 out
  • Relace$0.1259 in · $8 out
Intelligence71.2% correct#24 of 72
Speed56tok/s#57 of 72
Cost$0.0036per question#44 of 72List $1.32 in · $3.96 out per 1M
Verbosity1,318tokens per question#38 of 72

Summary

  • Answers 57 of 80 verification questions correctly (71.2%), placing #24 of 72 models tested on the current suite.
  • Generates 56 output tokens per second, faster than 21% of the models the leaderboard measures.
  • Starts responding after 0.8 s (median time to first token), including thinking time.
  • Costs $0.0036 per question as billed, #5 cheapest of 31 models priced over $1 per 1M tokens.

Specifications

Reasoning
Yes
Input
text
Output
text
Context window
1M tokens
Knowledge cutoff
—
OpenRouter ID
deepseek/deepseek-v4-pro-0813
Last tested
26 Oct 2026

Peer group: 34 models with a blended list price over $1 per 1M tokens (3 input : 1 output). Charts below compare DeepSeek V4 Pro 0813 with all models; cost and speed charts focus on this peer group.

Intelligence

Verification score: the share of 80 questions from open datasets answered correctly. See the benchmark datasets.

Score vs. cost per question

DeepSeek V4 Pro 0813 circled; dashed line joins models no cheaper model beats

Most attractive quadrant0255075100$0.00001$0.0001$0.001$0.01$0.10Cost per question (USD, log) →Verification score (%) →Qwen3.8 2.4T A95BCommand AReka Flash 3Gemma 4 26B A4BGemma 4 31BQwen3.5 397B A17BHermes 4 405BMuse WonderGrok 4.7o3gpt-oss-20bGranite 4.0 MicroNova Micro 1.0Mercury 2.5Phi 4Granite 4.2 8BHy3GLM 5.3 FlashSolar Pro 4Trinity Large ThinkingMiniMax M3Mistral Large 4Claude Opus 5.5GPT-6.1 SolGemini 4 ArgonGPT-5.4GPT-6 Lunagpt-oss-120b

Score vs. end-to-end response time

DeepSeek V4 Pro 0813 circled · accurate and quick sits top left

Most attractive quadrant02550751000 s25 s50 s75 s100 sEnd-to-end response time (s) →Verification score (%) →Qwen3.8 2.4T A95BCommand AReka Flash 3Llama 4 ScoutGemma 4 26B A4BLing 3.1 FlashHermes 4 405BMuse WonderGranite 4.0 MicroSolar Pro 4DeepSeek V4 Pro 0813Mistral Large 4Claude Opus 5.5Llama 4 MaverickKimi K3.1 Pro

Score vs. tokens per question

DeepSeek V4 Pro 0813 circled · output tokens, reasoning included

Most attractive quadrant02550751001001,00010,000Output tokens per question (log) →Verification score (%) →Command AReka Flash 3Llama 4 ScoutQwen3.7 PlusGemma 4 26B A4BGemma 4 31BQwen3.5 397B A17BHermes 4 405BClaude Fable 5.1GPT-6 AstraMuse WonderGrok 4.7o3gpt-oss-20bGranite 4.0 MicroNova Micro 1.0Mercury 2.5Granite 4.2 8BMinistral 3 14B 2512Hy3GLM 5.3 FlashSolar Pro 4Qwen3.6 35B A3BInkling SmallNemotron 3 UltraClaude Opus 5.5Gemini 4 ArgonGPT-5.4GPT-6 SolMiMo-V2.6-Pro

Verification score

All models tested on the current suite

  1. 1Claude Opus 5.585.0
  2. 2Gemini 4 Argon83.8
  3. 3GPT-5.3-Codex82.5
  4. 3Gemini 3.8 Flash82.5
  5. 5Claude Fable 5.181.2
  6. 6Grok 4.780.0
  7. 7Muse Wonder78.8
  8. 7Claude Sonnet 5.578.8
  9. 9Gemini 3.1 Pro Preview77.5
  10. 9Muse Spark Open77.5
  11. 24DeepSeek V4 Pro 081371.2

61 more models · see the full leaderboard

6 of 78 tested models have no verification score

Cost

Average cost of one question, split into input, reasoning and answer tokens at each model's list price. Reasoning tokens are billed as output.

Cost per question, by token type

Peer group: blended price over $1

  1. 1Inkling$0.0012
  2. 2GPT-6.1 Terra$0.0013
  3. 3GPT-6 Sol$0.0024
  4. 4GPT-6.1 Sol$0.0029
  5. 5Gemini 3.8 Flash$0.0041
  6. 6Muse Spark Open$0.0046
  7. 7Step 5 Preview$0.0047
  8. 8GPT-6.1 Terra Pro$0.005
  9. 9Nemotron 3 Ultra$0.0054
  10. 10Claude Sonnet 5.5$0.0054
  11. 11DeepSeek V4 Pro 0813$0.0056
  12. 12Kimi K2.7 Code$0.0057
  13. 13Gemini 4 Argon$0.0063
  14. 14GPT-5.6 Terra$0.0063
  15. 15Command A$0.007
  16. 16o3$0.0088
  17. 17Hermes 4 405B$0.01
  18. 18GPT-5.3-Codex$0.01
  19. 19Kimi K3.1 Pro$0.01
  20. 20Qwen3.8 Max (0902)$0.011
  21. 21Mistral Large 4$0.011
  22. 22GPT-5.4$0.011
  23. 23GPT-6 Astra$0.013
  24. 24Qwen3.5 397B A17B$0.015
  25. 25Claude Opus 5.5$0.015
  26. 26Qwen3.8 2.4T A95B$0.015
  27. 27Claude Fable 5.1$0.016
  28. 28Gemini 3.1 Pro Preview$0.017
  29. 29Muse Wonder$0.021
  30. 30Mistral Medium 3.5$0.025
  31. 31Grok 4.7$0.027
Input Reasoning Answer

Cost per question, as billed

All models · lower is better

  1. 1Granite 4.0 Micro$0.000051
  2. 2gpt-oss-20b$0.000066
  3. 3Granite 4.2 8B$0.000077
  4. 4Nova Micro 1.0$0.000087
  5. 5Mercury 2.5$0.00011
  6. 6Phi 4$0.00012
  7. 7gpt-oss-120b$0.00013
  8. 8GLM 5.3 Flash$0.00015
  9. 9Ministral 3 3B 2512$0.00019
  10. 10GPT-6 Luna$0.00025
  11. 44DeepSeek V4 Pro 0813$0.0036

61 more models · see the full leaderboard

6 of 78 tested models have no cost per question
  • LFM2.5-2.6B: no verification score (80 of 80 questions failed on the host's side)
  • Ling 3.0 Flash VL: no verification score (not yet run on the current question suite)
  • Qwen3.5-122B-A10B: no verification score (not yet run on the current question suite)
  • Qwen3 Next 80B A3B Thinking: no verification score (not yet run on the current question suite)
  • Nova Premier 1.0: no verification score (not yet run on the current question suite)
  • GPT-5.5 Pro: no verification score (not yet run on the current question suite)

Speed and latency

Median of 3 streamed runs of the same prompt, measured 25 Oct 2026.

Output speed

Peer group (over $1) · higher is better

  1. 1Qwen3.5 397B A17B240
  2. 2Muse Spark Open220
  3. 3Muse Wonder175
  4. 4o3163
  5. 5Mistral Medium 3.5145
  6. 6Gemini 3.8 Flash142
  7. 7Kimi K2.7 Code136
  8. 8Nemotron 3 Ultra136
  9. 9Claude Opus 5.5129
  10. 10Claude Sonnet 5.5126
  11. 11Gemini 3.1 Pro Preview124
  12. 12GPT-6.1 Terra122
  13. 23DeepSeek V4 Pro 081356

18 more models · see the full leaderboard

3 of 34 tested models have no output speed
  • GPT-6.1 Terra Pro: answer arrives in one burst after hidden thinking, so there is no real stream speed
  • Nova Premier 1.0: speed runs failed on the host's side
  • GPT-5.5 Pro: answer arrives in one burst after hidden thinking, so there is no real stream speed

Latency vs. output speed

DeepSeek V4 Pro 0813 circled · quick to start and fast sits top left

Most attractive quadrant02004006008000.0 s5.0 s10.0 s15.0 s20.0 sTime to first token (s) →Output speed (tok/s) →Nemotron 3.5 Super VLQwen3.5 397B A17BMuse WonderMinistral 3 3B 2512Trinity Large ThinkingGemini 3.5 Flash LiteGemini 3.1 Pro PreviewGemini 3.8 FlashGemini 4 Argon

Results by category

Questions answered correctly in each category

Science25/30
Engineering8/15
Medicine4/11
Economics8/9
Law4/6
Humanities & social science8/9

More from DeepSeek

DeepSeek#18DeepSeek V4.1 Flash 101773.8% verification score57 tok/s0.9 s first token$0.00091 / questionDeepSeek#24DeepSeek V4 Flash Vision Exp71.2% verification score170 tok/s0.7 s first token$0.0011 / question

All DeepSeek models →

Compare DeepSeek V4 Pro 0813

Head to head with its closest rivals, other DeepSeek models and the top models overall.

Hosts and prices

Every host serving this model through OpenRouter, from OpenRouter's catalogue

HostInput / 1MOutput / 1MContextMax outputQuantization
Baidu$0.13992$0.419761M393kfp8
Sail Research$0.45$0.81M384kfp4
Alibaba$0.5808$1.74241M393k—
StreamLake$0.66$1.981M384k—
DeepSeek$0.66$1.981M393k—
DeepInfra$1.3$2.61M16kfp8
Phala$0.957$2.87761M393k—
Ionstream$0.37$2.931M944k—
Novita$0.99$2.971M393kfp8
GMICloud$1.056$3.1681M944kfp8
DekaLLM$1.08$3.81M944k—
CoreWeave$1.31$3.961M944kfp8
BaseTen$1.32$3.961M262kfp4
AtlasCloud$1.32$3.961M393kfp8
Nebius$1.32$3.96979k881kfp8
Parasail$1.32$3.961M944kfp8
Together$1.32$3.961M944k—
DigitalOcean$1.32$3.961M944k—
SiliconFlow$1.32$3.961M393kfp8
Cloudflare$1.32$3.961M944k—
Wafer$0.285$4.31M944kmxfp4
Venice$1.65$4.951M33k—
Relace$0.1259$81M393kfp4

DeepSeek V4 Pro 0813: frequently asked questions

How accurate is DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 answered 57 of 80 questions correctly (71.2%) in the 2026-10-27 run, ranking #24 of 78 models.
How fast is DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 generated a median of 56 output tokens per second in the 2026-10-27 run, ranking #57 of 78. Median time to first token was 0.8 s.
What is the latency of DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 had a median time to first token of 0.8 s in the 2026-10-27 run, ranking #42 of 78 models.
How much does DeepSeek V4 Pro 0813 cost?
DeepSeek V4 Pro 0813 cost $0.0036 per question on average as billed in the 2026-10-27 run, reasoning tokens included.
What is DeepSeek V4 Pro 0813 API pricing?
DeepSeek V4 Pro 0813 is listed at $1.32 per million input tokens and $3.96 per million output tokens on OpenRouter.
What is the context window of DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 accepts up to 1,048,576 tokens of context, according to OpenRouter's model catalogue.
Is DeepSeek V4 Pro 0813 open source?
Yes. DeepSeek V4 Pro 0813 is an open-weights model: DeepSeek publishes the weights, so you can run it yourself as well as through an API.
Who makes DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 is made by DeepSeek. The LLM Verifications Leaderboard tested it through OpenRouter using the model ID deepseek/deepseek-v4-pro-0813.
Where can I use DeepSeek V4 Pro 0813 via API?
Through OpenRouter, DeepSeek V4 Pro 0813 is served by 12 hosts: Alibaba, Baidu, CoreWeave, DekaLLM, GMICloud, Ionstream, Phala, Relace, Sail Research, StreamLake, Together, Wafer.