DeepSeek-V4-Flash vs DeepSeek-V4-Pro 0813
Wins 0 of 7 areas
—
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Following instructions
DeepSeek-V4-Pro 0813 is the stronger all-rounder.DeepSeek-V4-Flash is cheaper.
Scores updated · 39 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking03DeepSeek-V4-Pro 08133 of 5 tests · 2 ties
- MathCompetition and research-level math03DeepSeek-V4-Pro 08133 of 3 tests
- CodingWriting and fixing software02DeepSeek-V4-Pro 08132 of 4 tests · 2 ties
- AgentsCarrying out multi-step tasks on its own01DeepSeek-V4-Pro 08131 of 3 tests · 2 ties
- FactsGetting facts right instead of making them up12DeepSeek-V4-Pro 08132 of 3 tests
- Following instructionsDoing exactly what it is asked01DeepSeek-V4-Pro 08131 of 1 test
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-V4-Flash costs 67% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where DeepSeek-V4-Pro 0813 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+18.8points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+8.7points ahead
- Fresh competition math problemsLiveBench · Mathematics+8.3points ahead
Where DeepSeek-V4-Flash pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+3.1points ahead
Every test, side by side
All 39 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro 0813
- LiveBench · Agentic CodingDeepSeek-V4-Pro 0813 by 8.246.855+8.2
- LiveBench · CodingDeepSeek-V4-Pro 0813 by 2.27577.2+2.2
- SciCodetie50.351tie
- Terminal-Bench 2.1tie78.778.7tie
AgentsDeepSeek-V4-Pro 0813
- Terminal-Bench 4.0DeepSeek-V4-Pro 0813 by 212.114.1+2
- GDPValtie46.747.5tie
- τ-Bench V3 · Bankingtie39.439.6tie
ReasoningDeepSeek-V4-Pro 0813
- Humanity's Last ExamDeepSeek-V4-Pro 0813 by 2.438.641+2.4
- GPQA DiamondDeepSeek-V4-Pro 0813 by 290.892.8+2
- CritPtDeepSeek-V4-Pro 0813 by 1.416.618+1.4
- LiveBench · Reasoningtie86.685.8tie
- ARC-AGI-2tie61.461.3tie
FactsDeepSeek-V4-Pro 0813
- SimpleQA VerifiedDeepSeek-V4-Pro 0813 by 18.834.152.9+18.8
- AA-Omniscience · AccuracyDeepSeek-V4-Pro 0813 by 8.740.449.1+8.7
- AA-Omniscience · Non-hallucinationDeepSeek-V4-Flash by 3.18.35.2+3.1
MathDeepSeek-V4-Pro 0813
- LiveBench · MathematicsDeepSeek-V4-Pro 0813 by 8.386.895.1+8.3
- FrontierMath Tiers 1-3 (v2)DeepSeek-V4-Pro 0813 by 7.157.564.6+7.1
- FrontierMath Tier 4DeepSeek-V4-Pro 0813 by 2.424.426.8+2.4
Following instructionsDeepSeek-V4-Pro 0813
- LiveBench · Instruction FollowingDeepSeek-V4-Pro 0813 by 2.265.567.7+2.2
Other results19 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- GDPval-AA v2DeepSeek-V4-Pro 0813 by 401 rating points11891590+401 rating
- AA-OmniscienceDeepSeek-V4-Pro 0813 by 15.1-14.30.8+15.1
- HLE (with tools)DeepSeek-V4-Pro 0813 by 14.945.160+14.9
- HLE (wo / w tools)DeepSeek-V4-Pro 0813 by 14.945.160+14.9
- DeepSWEDeepSeek-V4-Pro 0813 by 8.354.462.7+8.3
- DSBench-HardDeepSeek-V4-Pro 0813 by 7.659.667.2+7.6
- NL2RepoDeepSeek-V4-Pro 0813 by 7.354.261.5+7.3
- AutomationBench PublicDeepSeek-V4-Pro 0813 by 6.725.131.8+6.7
- CyberGymDeepSeek-V4-Pro 0813 by 6.676.783.3+6.6
- ToolathlonDeepSeek-V4-Pro 0813 by 3.870.374.1+3.8
- Toolathlon VerifiedDeepSeek-V4-Pro 0813 by 3.870.374.1+3.8
- livebench_languageDeepSeek-V4-Pro 0813 by 2.979.282.1+2.9
- DSBench-FullStackDeepSeek-V4-Pro 0813 by 2.468.771.1+2.4
- AA IntelligenceDeepSeek-V4-Pro 0813 by 1.734.336+1.7
- ARC-AGI-1DeepSeek-V4-Pro 0813 by 18990+1
- AA Agentic Indextie41.742.3tie
- Agents' Last Examtie25.225.7tie
- Artificial Analysis Coding Indextie69.168.8tie
- livebench_data_analysistie79.379.2tie
Questions people ask
Which is better, DeepSeek-V4-Flash or DeepSeek-V4-Pro 0813?
DeepSeek-V4-Pro 0813 wins six of the seven areas where both have results: coding, agents, reasoning, facts, math and following instructions. DeepSeek-V4-Flash wins none, but costs 67% less. They are level on long documents.
Which is better for coding?
DeepSeek-V4-Pro 0813. It wins 2 of the 4 coding tests both models report; DeepSeek-V4-Flash wins none, and 2 are ties.
Which is cheaper?
DeepSeek-V4-Flash costs $0.44 per million input tokens and $1.32 per million output tokens; DeepSeek-V4-Pro 0813 costs $1.32 and $3.96. That makes DeepSeek-V4-Flash about 67% cheaper for the same work.
How do you compare the two?
We use the 39 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.