Claude Opus 4.5 vs DeepSeek-V4-Pro
Wins 2 of 7 areas
Facts · Long documents
Wins 5 of 7 areas
Coding · Agents · Reasoning · Math · Following instructions
DeepSeek-V4-Pro is the stronger all-rounder.Claude Opus 4.5 is better at facts.
Scores updated · 48 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 software15DeepSeek-V4-Pro5 of 9 tests · 3 ties
- ReasoningHard problems that need careful thinking04DeepSeek-V4-Pro4 of 4 tests
- MathCompetition and research-level math14DeepSeek-V4-Pro4 of 6 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own12DeepSeek-V4-Pro2 of 3 tests
- Following instructionsDoing exactly what it is asked01DeepSeek-V4-Pro1 of 2 tests · 1 tie
- FactsGetting facts right instead of making them up21Claude Opus 4.52 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts10Claude Opus 4.51 of 1 test
Long documents rests on a single test.
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 96% 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
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+69.2points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+46.4points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+18.5points ahead
Where Claude Opus 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+33.1points ahead
- Real work tasks from 44 professionsGDPVal+14.4points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+9.7points ahead
Every test, side by side
All 48 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V4-Pro
- SWE-bench MultilingualDeepSeek-V4-Pro by 69.2776.2+69.2
- LiveBench · CodingClaude Opus 4.5 by 9.779.770+9.7
- LiveCodeBench v6DeepSeek-V4-Pro by 7.784.892.5+7.7
- SWE-bench ProDeepSeek-V4-Pro by 3.45255.4+3.4
- LiveBench · Agentic CodingDeepSeek-V4-Pro by 2.939.742.6+2.9
- SciCodeDeepSeek-V4-Pro by 1.349.550.8+1.3
- Terminal-Bench Hardtie4746.2tie
- LMArena · WebDevtie14691464tie
- SWE-bench Verifiedtie80.980.6tie
AgentsDeepSeek-V4-Pro
- BrowseCompDeepSeek-V4-Pro by 46.43783.4+46.4
- GDPValClaude Opus 4.5 by 14.447.332.9+14.4
- MCP AtlasDeepSeek-V4-Pro by 11.362.373.6+11.3
ReasoningDeepSeek-V4-Pro
- CritPtDeepSeek-V4-Pro by 8.34.612.9+8.3
- Humanity's Last ExamDeepSeek-V4-Pro by 7.430.137.5+7.4
- LiveBench · ReasoningDeepSeek-V4-Pro by 2.680.182.7+2.6
- GPQA DiamondDeepSeek-V4-Pro by 2.286.688.8+2.2
FactsClaude Opus 4.5
- AA-Omniscience · Non-hallucinationClaude Opus 4.5 by 33.1395.9+33.1
- AA-Omniscience · AccuracyClaude Opus 4.5 by 3.646.643+3.6
- Vectara HHEM hallucination ratelower is betterDeepSeek-V4-Pro by 2.310.98.6+2.3
- SimpleQA Verifiedtie45.746.2tie
MathDeepSeek-V4-Pro
- FrontierMath Tiers 1-3 (v2)DeepSeek-V4-Pro by 10.934.445.3+10.9
- HMMT Feb. 2026DeepSeek-V4-Pro by 9.985.395.2+9.9
- IMOAnswerBenchDeepSeek-V4-Pro by 5.88489.8+5.8
- FrontierMath Tier 4Claude Opus 4.5 by 2.54.92.4+2.5
- AIME 2026DeepSeek-V4-Pro by 1.695.196.7+1.6
- LiveBench · Mathematicstie90.490.7tie
Following instructionsDeepSeek-V4-Pro
- IFBenchDeepSeek-V4-Pro by 18.55876.5+18.5
- LiveBench · Instruction Followingtie62.562.4tie
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.
- CyberGymDeepSeek-V4-Pro by 32.750.683.3+32.7
- AA Agentic IndexClaude Opus 4.5 by 31.959.627.7+31.9
- AA-OmniscienceClaude Opus 4.5 by 24.714-10.7+24.7
- HLE (with tools)DeepSeek-V4-Pro by 16.843.260+16.8
- Artificial Analysis Coding IndexDeepSeek-V4-Pro by 11.647.859.4+11.6
- Terminal-Bench 2.0DeepSeek-V4-Pro by 8.659.367.9+8.6
- ToolathlonDeepSeek-V4-Pro by 8.343.551.8+8.3
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 6.789.596.2+6.7
- LiveCodeBenchDeepSeek-V4-Pro by 6.58793.5+6.5
- NL2RepoClaude Opus 4.5 by 4.743.238.5+4.7
- vectara_avg_summary_lengthDeepSeek-V4-Pro by 39.3 rating points114.5153.8+39.3 rating
- livebench_languageClaude Opus 4.5 by 3.281.378.1+3.2
- LiveBenchClaude Opus 4.5 by 2.47673.6+2.4
- vectara_factual_consistencyDeepSeek-V4-Pro by 2.389.191.4+2.3
- MMLU-ProClaude Opus 4.5 by 289.587.5+2
- vectara_answer_rateClaude Opus 4.5 by 1.598.797.2+1.5
- AA IntelligenceDeepSeek-V4-Pro by 1.329.130.4+1.3
- HMMT Nov. 2025DeepSeek-V4-Pro by 1.193.394.4+1.1
- livebench_data_analysistie74.474.5tie
Questions people ask
Which is better, Claude Opus 4.5 or DeepSeek-V4-Pro?
DeepSeek-V4-Pro wins five of the seven areas where both have results: coding, agents, reasoning, math and following instructions. Claude Opus 4.5 wins facts and long documents.
Which is better for coding?
DeepSeek-V4-Pro. It wins 5 of the 9 coding tests both models report; Claude Opus 4.5 wins 1, and 3 are ties.
Which is cheaper?
Claude Opus 4.5 costs $5.00 per million input tokens and $25.00 per million output tokens; DeepSeek-V4-Pro costs $0.43 and $0.87. That makes DeepSeek-V4-Pro about 96% cheaper for the same work.
How do you compare the two?
We use the 48 benchmark tests both models have published scores on. The verdict counts the 29 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.