Claude Opus 4.8 vs DeepSeek-V3.2
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Wins 0 of 7 areas
—
Claude Opus 4.8 is the stronger all-rounder.DeepSeek-V3.2 is cheaper.
Scores updated · 41 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 software70Claude Opus 4.87 of 7 tests
- AgentsCarrying out multi-step tasks on its own50Claude Opus 4.85 of 5 tests
- ReasoningHard problems that need careful thinking40Claude Opus 4.84 of 4 tests
- FactsGetting facts right instead of making them up30Claude Opus 4.83 of 3 tests
- MathCompetition and research-level math30Claude Opus 4.83 of 3 tests
- Long documentsFinding answers in very long texts10Claude Opus 4.81 of 1 test
- Following instructionsDoing exactly what it is asked10Claude Opus 4.81 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 98% 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 Claude Opus 4.8 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+68.1points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+53.6points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+43.4points ahead
Where DeepSeek-V3.2 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 41 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.8
- SWE-bench ProClaude Opus 4.8 by 53.669.215.6+53.6
- Terminal-Bench 2.1Claude Opus 4.8 by 37.884.646.8+37.8
- Terminal-Bench HardClaude Opus 4.8 by 22.758.335.6+22.7
- LMArena · WebDevClaude Opus 4.8 by 194 rating points15561362+194 rating
- SciCodeClaude Opus 4.8 by 15.554.438.9+15.5
- SWE-bench VerifiedClaude Opus 4.8 by 15.588.673.1+15.5
- SWE-bench MultilingualClaude Opus 4.8 by 14.284.470.2+14.2
AgentsClaude Opus 4.8
- GDPValClaude Opus 4.8 by 3847.89.8+38
- BrowseCompClaude Opus 4.8 by 32.984.351.4+32.9
- AA ApexAgentsClaude Opus 4.8 by 24.939.414.5+24.9
- MCP AtlasClaude Opus 4.8 by 2082.262.2+20
- τ-Bench V3 · BankingClaude Opus 4.8 by 15.434.218.8+15.4
ReasoningClaude Opus 4.8
- ARC-AGI-2Claude Opus 4.8 by 68.172.14+68.1
- Humanity's Last ExamClaude Opus 4.8 by 24.148.724.6+24.1
- CritPtClaude Opus 4.8 by 1820.92.9+18
- GPQA DiamondClaude Opus 4.8 by 89284+8
FactsClaude Opus 4.8
- AA-Omniscience · Non-hallucinationClaude Opus 4.8 by 43.460.717.3+43.4
- SimpleQA VerifiedClaude Opus 4.8 by 25.55327.5+25.5
- AA-Omniscience · AccuracyClaude Opus 4.8 by 15.848.833+15.8
MathClaude Opus 4.8
- HMMT Feb. 2026Claude Opus 4.8 by 11.495.584.1+11.4
- AIME 2026Claude Opus 4.8 by 5.810094.2+5.8
- IMOAnswerBenchClaude Opus 4.8 by 5.283.578.3+5.2
Following instructionsClaude Opus 4.8
- IFBenchClaude Opus 4.8 by 1.562.260.7+1.5
Other results17 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.
- CyberGymClaude Opus 4.8 by 61.578.817.3+61.5
- AA-OmniscienceClaude Opus 4.8 by 51.328.8-22.5+51.3
- ARC-AGI-1Claude Opus 4.8 by 35.592.557+35.5
- PaperBenchClaude Opus 4.8 by 33.280.347.1+33.2
- DeepSearchQA (F1)Claude Opus 4.8 by 32.293.160.9+32.2
- Artificial Analysis Coding IndexClaude Opus 4.8 by 30.174.344.2+30.1
- frontiermath_tier_4_v1Claude Opus 4.8 by 29.231.32.1+29.2
- Terminal-Bench 2.0Claude Opus 4.8 by 28.274.646.4+28.2
- Tool-DecathlonClaude Opus 4.8 by 24.759.935.2+24.7
- ToolathlonClaude Opus 4.8 by 24.759.935.2+24.7
- AA Agentic IndexClaude Opus 4.8 by 24.342.618.3+24.3
- AA IntelligenceClaude Opus 4.8 by 20.341.821.5+20.3
- ResearchRubricsClaude Opus 4.8 by 17.773.555.8+17.7
- HLE (with tools)Claude Opus 4.8 by 17.157.940.8+17.1
- HealthBenchClaude Opus 4.8 by 12.459.346.9+12.4
- HMMT Nov. 2025Claude Opus 4.8 by 6.596.590+6.5
- τ²-Bench Telecom (AA run)Claude Opus 4.8 by 3.894.490.6+3.8
Questions people ask
Which is better, Claude Opus 4.8 or DeepSeek-V3.2?
Claude Opus 4.8 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. DeepSeek-V3.2 wins none, but costs 98% less.
Which is better for coding?
Claude Opus 4.8. It wins 7 of the 7 coding tests both models report; DeepSeek-V3.2 wins none.
Which is cheaper?
Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens; DeepSeek-V3.2 costs $0.28 and $0.42. That makes DeepSeek-V3.2 about 98% cheaper for the same work.
How do you compare the two?
We use the 41 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 17 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.