DeepSeek-V4-Pro vs GPT-5
Wins 4 of 7 areas
Coding · Agents · Reasoning · Following instructions
Wins 2 of 7 areas
Math · Long documents
DeepSeek-V4-Pro wins more areas, narrowly.GPT-5 is better at math.
Scores updated · 38 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.
- CodingWriting and fixing software80DeepSeek-V4-Pro8 of 8 tests
- AgentsCarrying out multi-step tasks on its own30DeepSeek-V4-Pro3 of 3 tests
- ReasoningHard problems that need careful thinking31DeepSeek-V4-Pro3 of 4 tests
- Following instructionsDoing exactly what it is asked10DeepSeek-V4-Pro1 of 1 test
- MathCompetition and research-level math12GPT-52 of 3 tests
- Long documentsFinding answers in very long texts01GPT-51 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
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-Pro costs 88% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where DeepSeek-V4-Pro pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+28.8points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+28.5points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+20.9points ahead
Where GPT-5 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+19.6points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+11.9points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+10.1points ahead
Every test, side by side
All 38 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro
- Terminal-Bench 2.1DeepSeek-V4-Pro by 28.86435.2+28.8
- SWE-bench MultilingualDeepSeek-V4-Pro by 20.976.255.3+20.9
- SWE-bench ProDeepSeek-V4-Pro by 13.655.441.8+13.6
- Terminal-Bench HardDeepSeek-V4-Pro by 13.646.232.6+13.6
- SciCodeDeepSeek-V4-Pro by 7.950.842.9+7.9
- SWE-bench VerifiedDeepSeek-V4-Pro by 5.780.674.9+5.7
- LiveCodeBench v6DeepSeek-V4-Pro by 5.592.587+5.5
- LMArena · WebDevDeepSeek-V4-Pro by 46 rating points14641418+46 rating
AgentsDeepSeek-V4-Pro
- BrowseCompDeepSeek-V4-Pro by 28.583.454.9+28.5
- GDPValDeepSeek-V4-Pro by 11.932.921+11.9
- τ-Bench V3 · BankingDeepSeek-V4-Pro by 830.122.1+8
ReasoningDeepSeek-V4-Pro
- Humanity's Last ExamDeepSeek-V4-Pro by 937.528.5+9
- CritPtDeepSeek-V4-Pro by 7.212.95.7+7.2
- SimpleBenchGPT-5 by 5.850.956.7+5.8
- GPQA DiamondDeepSeek-V4-Pro by 3.488.885.4+3.4
FactsEven
- AA-Omniscience · Non-hallucinationGPT-5 by 11.95.917.8+11.9
- Vectara HHEM hallucination ratelower is betterDeepSeek-V4-Pro by 6.58.615.1+6.5
- SimpleQA VerifiedGPT-5 by 3.946.250.1+3.9
- AA-Omniscience · AccuracyDeepSeek-V4-Pro by 2.74340.3+2.7
MathGPT-5
- FrontierMath Tier 4GPT-5 by 19.62.421.9+19.6
- IMOAnswerBenchDeepSeek-V4-Pro by 13.889.876+13.8
- FrontierMath Tiers 1-3 (v2)GPT-5 by 10.145.355.4+10.1
Following instructionsDeepSeek-V4-Pro
- IFBenchDeepSeek-V4-Pro by 3.476.573.1+3.4
Other results14 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.
- Terminal-Bench 2.0DeepSeek-V4-Pro by 32.767.935.2+32.7
- HLE (with tools)DeepSeek-V4-Pro by 24.86035.2+24.8
- Artificial Analysis Coding IndexDeepSeek-V4-Pro by 21.659.437.8+21.6
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 11.496.284.8+11.4
- LiveCodeBenchDeepSeek-V4-Pro by 1093.583.5+10
- AA IntelligenceDeepSeek-V4-Pro by 7.430.423+7.4
- vectara_factual_consistencyDeepSeek-V4-Pro by 6.591.484.9+6.5
- HMMT Nov. 2025DeepSeek-V4-Pro by 5.294.489.2+5.2
- vectara_answer_rateGPT-5 by 2.797.299.9+2.7
- GPQA (unspecified)DeepSeek-V4-Pro by 2.187.885.7+2.1
- AA-OmniscienceGPT-5 by 2-10.7-8.7+2
- AA Agentic IndexDeepSeek-V4-Pro by 1.227.726.5+1.2
- vectara_avg_summary_lengthtie153.8162.7tie
- MMLU-Protie87.587.1tie
Questions people ask
Which is better, DeepSeek-V4-Pro or GPT-5?
DeepSeek-V4-Pro wins four of the seven areas where both have results: coding, agents, reasoning and following instructions. GPT-5 wins math and long documents. They are level on facts.
Which is better for coding?
DeepSeek-V4-Pro. It wins 8 of the 8 coding tests both models report; GPT-5 wins none.
Which is cheaper?
DeepSeek-V4-Pro costs $0.43 per million input tokens and $0.87 per million output tokens; GPT-5 costs $1.25 and $10.00. That makes DeepSeek-V4-Pro about 88% cheaper for the same work.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 24 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 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.