GPT-5.4 vs Claude Opus 5
Wins 2 of 8 areas
Long documents · Following instructions
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Claude Opus 5 is the stronger all-rounder.GPT-5.4 is cheaper and better at long documents.
Scores updated · 45 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. 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 software06Claude Opus 56 of 7 tests · 1 tie
- ReasoningHard problems that need careful thinking06Claude Opus 56 of 6 tests
- AgentsCarrying out multi-step tasks on its own05Claude Opus 55 of 5 tests
- FactsGetting facts right instead of making them up03Claude Opus 53 of 3 tests
- MathCompetition and research-level math03Claude Opus 53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video02Claude Opus 52 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts10GPT-5.41 of 1 test
- Following instructionsDoing exactly what it is asked10GPT-5.41 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.
GPT-5.4 costs 42% 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 5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+30.9points ahead
- Solves new puzzle games it has never seenARC-AGI-3+30points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+24.2points ahead
Where GPT-5.4 pulls ahead
- Follows detailed instructions exactlyLiveBench · Instruction Following+6.4points ahead
- Reasons across sets of long documentsAA-LCR+2.7points ahead
Every test, side by side
All 45 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 5
- LMArena · WebDevClaude Opus 5 by 227 rating points14651692+227 rating
- SWE-bench ProClaude Opus 5 by 21.557.779.2+21.5
- SWE-bench MultilingualClaude Opus 5 by 17.871.789.5+17.8
- LiveBench · Agentic CodingClaude Opus 5 by 11.453.865.2+11.4
- Terminal-Bench 2.1Claude Opus 5 by 10.878.389.1+10.8
- LiveBench · CodingClaude Opus 5 by 477.581.5+4
- SciCodetie56.656.4tie
AgentsClaude Opus 5
- GDPValClaude Opus 5 by 23.837.461.2+23.8
- MCP AtlasClaude Opus 5 by 18.667.285.8+18.6
- OSWorld-VerifiedClaude Opus 5 by 8.47583.4+8.4
- BrowseCompClaude Opus 5 by 8.182.790.8+8.1
- τ-Bench V3 · BankingClaude Opus 5 by 2.539.642.1+2.5
ReasoningClaude Opus 5
- ARC-AGI-3Claude Opus 5 by 300.230.2+30
- ARC-AGI-2Claude Opus 5 by 17.173.390.4+17.1
- Humanity's Last ExamClaude Opus 5 by 11.243.754.9+11.2
- CritPtClaude Opus 5 by 5.723.429.1+5.7
- LiveBench · ReasoningClaude Opus 5 by 3.188.191.2+3.1
- GPQA DiamondClaude Opus 5 by 1.29293.2+1.2
FactsClaude Opus 5
- AA-Omniscience · Non-hallucinationClaude Opus 5 by 30.98.339.2+30.9
- SimpleQA VerifiedClaude Opus 5 by 14.845.159.9+14.8
- AA-Omniscience · AccuracyClaude Opus 5 by 1050.960.9+10
Images and chartsClaude Opus 5
- MMMU-ProClaude Opus 5 by 6.378.484.7+6.3
- LMArena · VisionClaude Opus 5 by 19 rating points13021321+19 rating
- CharXiv (RQ)tie82.883.7tie
MathClaude Opus 5
- FrontierMath Tier 4Claude Opus 5 by 24.24973.2+24.2
- FrontierMath Tiers 1-3 (v2)Claude Opus 5 by 778.685.6+7
- LiveBench · MathematicsClaude Opus 5 by 1.594.295.7+1.5
Following instructionsGPT-5.4
- LiveBench · Instruction FollowingGPT-5.4 by 6.470.263.8+6.4
Other results16 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.
- HiL-Bench (Tools-allowed)Claude Opus 5 by 47.39.757+47.3
- AA-OmniscienceClaude Opus 5 by 31.35.837.1+31.3
- DeepSWE 1.1Claude Opus 5 by 16.85268.8+16.8
- OfficeQA ProClaude Opus 5 by 15.851.166.9+15.8
- AA Agentic IndexClaude Opus 5 by 1244.256.2+12
- AA IntelligenceClaude Opus 5 by 11.83950.8+11.8
- HealthBench Professional length-adjustedClaude Opus 5 by 11.748.159.8+11.7
- HLE (with tools)Claude Opus 5 by 11.552.163.6+11.5
- Legal Agent BenchmarkClaude Opus 5 by 11.30.411.7+11.3
- OfficeQAClaude Opus 5 by 1068.178.1+10
- Artificial Analysis Coding IndexClaude Opus 5 by 6.971.178+6.9
- livebench_languageClaude Opus 5 by 6.182.688.7+6.1
- livebench_data_analysisGPT-5.4 by 4.779.374.5+4.7
- ARC-AGI-1Claude Opus 5 by 3.893.797.5+3.8
- HealthBench length-adjustedClaude Opus 5 by 3.85457.8+3.8
- GDPval-AA (Elo)Claude Opus 5 by 34 rating points16741708+34 rating
Questions people ask
Which is better, GPT-5.4 or Claude Opus 5?
Claude Opus 5 wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and math. GPT-5.4 wins long documents and following instructions, and costs 42% less.
Which is better for coding?
Claude Opus 5. It wins 6 of the 7 coding tests both models report; GPT-5.4 wins none, and 1 is a tie.
Which is cheaper?
GPT-5.4 costs $2.50 per million input tokens and $15.00 per million output tokens; Claude Opus 5 costs $5.00 and $25.00. That makes GPT-5.4 about 42% cheaper for the same work.
How do you compare the two?
We use the 45 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 16 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.