Claude Sonnet 4 vs DeepSeek-V3.2
Wins 0 of 6 areas
—
Wins 6 of 6 areas
Coding · Agents · Reasoning · Facts · Long documents · Following instructions
DeepSeek-V3.2 is the stronger all-rounder.
Scores updated · 44 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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 software24DeepSeek-V3.24 of 7 tests · 1 tie
- ReasoningHard problems that need careful thinking13DeepSeek-V3.23 of 4 tests
- AgentsCarrying out multi-step tasks on its own02DeepSeek-V3.22 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up12DeepSeek-V3.22 of 3 tests
- Long documentsFinding answers in very long texts01DeepSeek-V3.21 of 1 test
- Following instructionsDoing exactly what it is asked01DeepSeek-V3.21 of 1 test
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-V3.2 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-V3.2 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+39.2points ahead
- Recent programming contest problemsLiveCodeBench v6+34.8points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+13.9points ahead
Where Claude Sonnet 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+53.6points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+27.1points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+1.9points ahead
Every test, side by side
All 44 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.2
- LiveCodeBench v6DeepSeek-V3.2 by 34.848.583.3+34.8
- SWE-bench ProClaude Sonnet 4 by 27.142.715.6+27.1
- SWE-bench MultilingualDeepSeek-V3.2 by 13.356.970.2+13.3
- Terminal-Bench 2.1DeepSeek-V3.2 by 10.536.346.8+10.5
- Terminal-Bench HardDeepSeek-V3.2 by 4.531.135.6+4.5
- SciCodeClaude Sonnet 4 by 1.14038.9+1.1
- SWE-bench Verifiedtie72.773.1tie
AgentsDeepSeek-V3.2
- BrowseCompDeepSeek-V3.2 by 39.212.251.4+39.2
- τ-Bench V3 · BankingDeepSeek-V3.2 by 2.116.718.8+2.1
- GDPValtie9.79.8tie
ReasoningDeepSeek-V3.2
- Humanity's Last ExamDeepSeek-V3.2 by 13.910.724.6+13.9
- GPQA DiamondDeepSeek-V3.2 by 6.377.784+6.3
- CritPtDeepSeek-V3.2 by 2.60.32.9+2.6
- ARC-AGI-2Claude Sonnet 4 by 1.95.94+1.9
FactsDeepSeek-V3.2
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 53.670.917.3+53.6
- AA-Omniscience · AccuracyDeepSeek-V3.2 by 10.322.733+10.3
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.2 by 410.36.3+4
Following instructionsDeepSeek-V3.2
- IFBenchDeepSeek-V3.2 by 654.760.7+6
Other results25 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.
- HMMT 2025DeepSeek-V3.2 by 74.315.990.2+74.3
- BrowseComp-ZHDeepSeek-V3.2 by 35.929.165+35.9
- τ²-BenchDeepSeek-V3.2 by 35.145.280.3+35.1
- τ²-Bench Telecom (AA run)DeepSeek-V3.2 by 2664.690.6+26
- AA-OmniscienceClaude Sonnet 4 by 22.70.2-22.5+22.7
- AIME 2025DeepSeek-V3.2 by 22.670.593.1+22.6
- HLE (with tools)DeepSeek-V3.2 by 20.520.340.8+20.5
- ARC-AGI-1DeepSeek-V3.2 by 174057+17
- FinSearchComp-globalClaude Sonnet 4 by 15.84226.2+15.8
- LiveCodeBenchDeepSeek-V3.2 by 14.868.583.3+14.8
- xbench-DeepSearchClaude Sonnet 4 by 8.964.655.7+8.9
- vectara_avg_summary_lengthClaude Sonnet 4 by 83.8 rating points145.862+83.8 rating
- Artificial Analysis Coding IndexDeepSeek-V3.2 by 6.637.644.2+6.6
- vectara_answer_rateClaude Sonnet 4 by 698.692.6+6
- Multi-SWE-BenchClaude Sonnet 4 by 5.135.730.6+5.1
- GAIA (text only)Claude Sonnet 4 by 4.868.363.5+4.8
- vectara_factual_consistencyDeepSeek-V3.2 by 489.793.7+4
- theagentcompanyClaude Sonnet 4 by 33734+3
- AA IntelligenceDeepSeek-V3.2 by 2.618.921.5+2.6
- Terminal-BenchDeepSeek-V3.2 by 2.235.537.7+2.2
- frontiermath_tier_4_v1DeepSeek-V3.2 by 2.102.1+2.1
- ArtifactsBenchClaude Sonnet 4 by 1.557.355.8+1.5
- MMLU-ProDeepSeek-V3.2 by 1.383.785+1.3
- AA Agentic Indextie17.618.3tie
- MMLU-Reduxtie93.693.7tie
Questions people ask
Which is better, Claude Sonnet 4 or DeepSeek-V3.2?
DeepSeek-V3.2 wins all six areas where both have results: coding, agents, reasoning, facts, long documents and following instructions. Claude Sonnet 4 wins none.
Which is better for coding?
DeepSeek-V3.2. It wins 4 of the 7 coding tests both models report; Claude Sonnet 4 wins 2, and 1 is a tie.
Which is cheaper?
Claude Sonnet 4 costs $3.00 per million input tokens and $15.00 per million output tokens; DeepSeek-V3.2 costs $0.28 and $0.42. That makes DeepSeek-V3.2 about 96% cheaper for the same work.
How do you compare the two?
We use the 44 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 25 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.