Claude Opus 4.6 vs DeepSeek-V4-Pro
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents
Wins 1 of 7 areas
Following instructions
Claude Opus 4.6 is the stronger all-rounder.DeepSeek-V4-Pro is cheaper and better at following instructions.
Scores updated · 66 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 software52Claude Opus 4.65 of 9 tests · 2 ties
- ReasoningHard problems that need careful thinking30Claude Opus 4.63 of 5 tests · 2 ties
- MathCompetition and research-level math42Claude Opus 4.64 of 7 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own21Claude Opus 4.62 of 4 tests · 1 tie
- FactsGetting facts right instead of making them up21Claude Opus 4.62 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts10Claude Opus 4.61 of 1 test
- Following instructionsDoing exactly what it is asked01DeepSeek-V4-Pro1 of 2 tests · 1 tie
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 Claude Opus 4.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+31.3points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+24.4points ahead
- Real work tasks from 44 professionsGDPVal+23points ahead
Where DeepSeek-V4-Pro pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+23.4points ahead
- Olympiad math problems with short answersIMOAnswerBench+14.5points ahead
- Multi-step tasks using many real software toolsMCP Atlas+10.9points ahead
Every test, side by side
All 66 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.6
- LiveBench · CodingClaude Opus 4.6 by 8.278.270+8.2
- LMArena · WebDevClaude Opus 4.6 by 82 rating points15461464+82 rating
- LiveBench · Agentic CodingClaude Opus 4.6 by 6.44942.6+6.4
- LiveCodeBench v6DeepSeek-V4-Pro by 3.788.892.5+3.7
- SWE-bench ProDeepSeek-V4-Pro by 253.455.4+2
- SWE-bench MultilingualClaude Opus 4.6 by 1.677.876.2+1.6
- SciCodeClaude Opus 4.6 by 1.151.950.8+1.1
- SWE-bench Verifiedtie80.880.6tie
- Terminal-Bench Hardtie46.246.2tie
AgentsClaude Opus 4.6
- GDPValClaude Opus 4.6 by 2355.932.9+23
- MCP AtlasDeepSeek-V4-Pro by 10.962.773.6+10.9
- AA ApexAgentsClaude Opus 4.6 by 8.73324.3+8.7
- BrowseComptie8483.4tie
ReasoningClaude Opus 4.6
- SimpleBenchClaude Opus 4.6 by 16.767.650.9+16.7
- LiveBench · ReasoningClaude Opus 4.6 by 688.782.7+6
- Humanity's Last ExamClaude Opus 4.6 by 2.439.937.5+2.4
- GPQA Diamondtie89.688.8tie
- CritPttie12.612.9tie
FactsClaude Opus 4.6
- AA-Omniscience · Non-hallucinationClaude Opus 4.6 by 31.337.25.9+31.3
- AA-Omniscience · AccuracyClaude Opus 4.6 by 44743+4
- Vectara HHEM hallucination ratelower is betterDeepSeek-V4-Pro by 3.612.28.6+3.6
- SimpleQA Verifiedtie4746.2tie
MathClaude Opus 4.6
- FrontierMath Tier 4Claude Opus 4.6 by 24.426.82.4+24.4
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.6 by 20.76645.3+20.7
- IMOAnswerBenchDeepSeek-V4-Pro by 14.575.389.8+14.5
- USAMO 2026Claude Opus 4.6 by 5.566.260.7+5.5
- LiveBench · MathematicsDeepSeek-V4-Pro by 1.489.390.7+1.4
- HMMT Feb. 2026Claude Opus 4.6 by 196.295.2+1
- AIME 2026tie96.796.7tie
Following instructionsDeepSeek-V4-Pro
- IFBenchDeepSeek-V4-Pro by 23.453.176.5+23.4
- LiveBench · Instruction Followingtie63.362.4tie
Other results34 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.
- AA Agentic IndexClaude Opus 4.6 by 39.967.627.7+39.9
- AA-OmniscienceClaude Opus 4.6 by 24.413.7-10.7+24.4
- CorpusQA 1M (ACC)Claude Opus 4.6 by 15.271.756.5+15.2
- Artificial Analysis Coding IndexDeepSeek-V4-Pro by 11.348.159.4+11.3
- NL2RepoClaude Opus 4.6 by 11.349.838.5+11.3
- OpenAI-MRCR (1M)Claude Opus 4.6 by 9.692.983.3+9.6
- CyberGymDeepSeek-V4-Pro by 9.573.883.3+9.5
- Apex (Pass@1)Claude Opus 4.6 by 7.134.527.4+7.1
- HLE (with tools)DeepSeek-V4-Pro by 6.753.360+6.7
- AgentWorldBench - MCPClaude Opus 4.6 by 6.669.963.3+6.6
- AgentWorldBench - AndroidClaude Opus 4.6 by 6.561.755.2+6.5
- AgentWorldBench - OSClaude Opus 4.6 by 6.570.263.7+6.5
- GDPval-AA (Elo)Claude Opus 4.6 by 65 rating points16191554+65 rating
- AgentWorldBench - TerminalClaude Opus 4.6 by 6.257.551.3+6.2
- Tool-DecathlonDeepSeek-V4-Pro by 5.647.252.8+5.6
- AgentWorldBench - SWEClaude Opus 4.6 by 5.264.559.4+5.2
- livebench_languageClaude Opus 4.6 by 5.283.378.1+5.2
- ToolathlonClaude Opus 4.6 by 556.851.8+5
- AgentWorldBench - OverallClaude Opus 4.6 by 4.857.853+4.8
- LiveCodeBenchDeepSeek-V4-Pro by 4.788.893.5+4.7
- livebench_data_analysisDeepSeek-V4-Pro by 4.669.974.5+4.6
- τ²-Bench Telecom (AA run)DeepSeek-V4-Pro by 4.192.196.2+4.1
- vectara_factual_consistencyDeepSeek-V4-Pro by 3.687.891.4+3.6
- LiveBenchClaude Opus 4.6 by 2.776.373.6+2.7
- vectara_answer_rateClaude Opus 4.6 by 2.699.897.2+2.6
- Terminal-Bench 2.0DeepSeek-V4-Pro by 2.565.467.9+2.5
- HMMT Nov. 2025Claude Opus 4.6 by 1.996.394.4+1.9
- AgentWorldBench - SearchClaude Opus 4.6 by 1.729.327.6+1.7
- vectara_avg_summary_lengthDeepSeek-V4-Pro by 16.2 rating points137.6153.8+16.2 rating
- MMLU-ProClaude Opus 4.6 by 1.689.187.5+1.6
- AA IntelligenceClaude Opus 4.6 by 1.531.930.4+1.5
- AgentWorldBench - WebClaude Opus 4.6 by 1.151.450.3+1.1
- Apex Shortlist (Pass@1)tie85.985.5tie
- MCPAtlas Public (Pass@1)tie73.873.6tie
Questions people ask
Which is better, Claude Opus 4.6 or DeepSeek-V4-Pro?
Claude Opus 4.6 wins six of the seven areas where both have results: coding, agents, reasoning, facts, math and long documents. DeepSeek-V4-Pro wins following instructions, and costs 96% less.
Which is better for coding?
Claude Opus 4.6. It wins 5 of the 9 coding tests both models report; DeepSeek-V4-Pro wins 2, and 2 are ties.
Which is cheaper?
Claude Opus 4.6 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 66 benchmark tests both models have published scores on. The verdict counts the 32 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 34 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.