Claude Opus 4.5 vs GPT-5.4
Wins 0 of 8 areas
—
Wins 8 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents · Following instructions
GPT-5.4 is the stronger all-rounder.
Scores updated · 55 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.
- MathCompetition and research-level math06GPT-5.46 of 6 tests
- ReasoningHard problems that need careful thinking05GPT-5.45 of 5 tests
- CodingWriting and fixing software15GPT-5.45 of 7 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own13GPT-5.43 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02GPT-5.42 of 2 tests
- Following instructionsDoing exactly what it is asked02GPT-5.42 of 2 tests
- FactsGetting facts right instead of making them up12GPT-5.42 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts01GPT-5.41 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.
GPT-5.4 costs 42% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.4 pulls ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+64.7points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+45.7points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+44.2points ahead
Where Claude Opus 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+30.7points ahead
- Real work tasks from 44 professionsGDPVal+9.9points ahead
- Programming problems, refreshed regularlyLiveBench · Coding+2.2points ahead
Every test, side by side
All 55 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.4
- SWE-bench MultilingualGPT-5.4 by 64.7771.7+64.7
- LiveBench · Agentic CodingGPT-5.4 by 14.139.753.8+14.1
- Terminal-Bench HardGPT-5.4 by 10.64757.6+10.6
- SciCodeGPT-5.4 by 7.149.556.6+7.1
- SWE-bench ProGPT-5.4 by 5.75257.7+5.7
- LiveBench · CodingClaude Opus 4.5 by 2.279.777.5+2.2
- LMArena · WebDevtie14691465tie
AgentsGPT-5.4
- BrowseCompGPT-5.4 by 45.73782.7+45.7
- GDPValClaude Opus 4.5 by 9.947.337.4+9.9
- OSWorld-VerifiedGPT-5.4 by 8.766.375+8.7
- MCP AtlasGPT-5.4 by 4.962.367.2+4.9
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 35.737.673.3+35.7
- CritPtGPT-5.4 by 18.84.623.4+18.8
- Humanity's Last ExamGPT-5.4 by 13.630.143.7+13.6
- LiveBench · ReasoningGPT-5.4 by 880.188.1+8
- GPQA DiamondGPT-5.4 by 5.486.692+5.4
FactsGPT-5.4
- AA-Omniscience · Non-hallucinationClaude Opus 4.5 by 30.7398.3+30.7
- AA-Omniscience · AccuracyGPT-5.4 by 4.346.650.9+4.3
- Vectara HHEM hallucination ratelower is betterGPT-5.4 by 3.910.97+3.9
- SimpleQA Verifiedtie45.745.1tie
Images and chartsGPT-5.4
- CharXiv (RQ)GPT-5.4 by 15.667.282.8+15.6
- MMMU-ProGPT-5.4 by 4.47478.4+4.4
MathGPT-5.4
- FrontierMath Tiers 1-3 (v2)GPT-5.4 by 44.234.478.6+44.2
- FrontierMath Tier 4GPT-5.4 by 44.14.949+44.1
- HMMT Feb. 2026GPT-5.4 by 12.485.397.7+12.4
- IMOAnswerBenchGPT-5.4 by 7.48491.4+7.4
- AIME 2026GPT-5.4 by 4.195.199.2+4.1
- LiveBench · MathematicsGPT-5.4 by 3.890.494.2+3.8
Following instructionsGPT-5.4
- IFBenchGPT-5.4 by 15.95873.9+15.9
- LiveBench · Instruction FollowingGPT-5.4 by 7.662.570.2+7.6
Other results24 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.
- Artificial Analysis Coding IndexGPT-5.4 by 23.347.871.1+23.3
- frontiermath_tier_4_v1GPT-5.4 by 22.94.227.1+22.9
- Terminal-Bench 2.0GPT-5.4 by 15.859.375.1+15.8
- CyberGymGPT-5.4 by 15.750.666.3+15.7
- AA Agentic IndexClaude Opus 4.5 by 15.459.644.2+15.4
- MATH-VisionGPT-5.4 by 14.977.192+14.9
- ARC-AGI-1GPT-5.4 by 11.78293.7+11.7
- ToolathlonGPT-5.4 by 11.143.554.6+11.1
- AA IntelligenceGPT-5.4 by 9.929.139+9.9
- HLE (with tools)GPT-5.4 by 8.843.252+8.8
- AA-OmniscienceClaude Opus 4.5 by 8.2145.8+8.2
- livebench_data_analysisGPT-5.4 by 4.974.479.3+4.9
- LiveBenchGPT-5.4 by 4.37680.3+4.3
- vectara_factual_consistencyGPT-5.4 by 3.989.193+3.9
- vectara_avg_summary_lengthClaude Opus 4.5 by 32.8 rating points114.581.7+32.8 rating
- DeepSearchQA (F1)GPT-5.4 by 2.576.178.6+2.5
- HMMT Nov. 2025GPT-5.4 by 2.593.395.8+2.5
- τ²-Bench Telecom (AA run)Claude Opus 4.5 by 2.489.587.1+2.4
- MMLU-ProClaude Opus 4.5 by 289.587.5+2
- NL2RepoClaude Opus 4.5 by 1.943.241.3+1.9
- OmniDocBench 1.5GPT-5.4 by 1.487.789.1+1.4
- livebench_languageGPT-5.4 by 1.381.382.6+1.3
- vectara_answer_rateGPT-5.4 by 1.298.799.9+1.2
- Finance Agenttie55.256tie
Questions people ask
Which is better, Claude Opus 4.5 or GPT-5.4?
GPT-5.4 wins all eight areas we test: coding, agents, reasoning, facts, images and charts, math, long documents and following instructions. Claude Opus 4.5 wins none.
Which is better for coding?
GPT-5.4. It wins 5 of the 7 coding tests both models report; Claude Opus 4.5 wins 1, and 1 is a tie.
Which is cheaper?
Claude Opus 4.5 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5.4 costs $2.50 and $15.00. That makes GPT-5.4 about 42% cheaper for the same work.
How do you compare the two?
We use the 55 benchmark tests both models have published scores on. The verdict counts the 31 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.