Claude Opus 4.6 vs GPT-5
Wins 5 of 8 areas
Coding · Agents · Reasoning · Facts · Math
Wins 1 of 8 areas
Following instructions
Claude Opus 4.6 is the stronger all-rounder.GPT-5 is cheaper and better at following instructions.
Scores updated · 48 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 software70Claude Opus 4.67 of 7 tests
- ReasoningHard problems that need careful thinking50Claude Opus 4.65 of 5 tests
- FactsGetting facts right instead of making them up31Claude Opus 4.63 of 4 tests
- MathCompetition and research-level math20Claude Opus 4.62 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own20Claude Opus 4.62 of 2 tests
- Following instructionsDoing exactly what it is asked02GPT-52 of 2 tests
- Images and chartsUnderstanding pictures, charts and video22Even2 each
- Long documentsFinding answers in very long texts00Even0 each · 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.
GPT-5 costs 63% 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
- Abstract visual puzzles that people can solveARC-AGI-2+58.9points ahead
- Real work tasks from 44 professionsGDPVal+34.9points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+29.1points ahead
Where GPT-5 pulls ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+32.4points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+20points ahead
- Reads and reasons about text in imagesOCRBenchv2+7.1points ahead
Every test, side by side
All 48 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.6
- SWE-bench MultilingualClaude Opus 4.6 by 22.577.855.3+22.5
- Terminal-Bench HardClaude Opus 4.6 by 13.646.232.6+13.6
- LMArena · WebDevClaude Opus 4.6 by 128 rating points15461418+128 rating
- SWE-bench ProClaude Opus 4.6 by 11.653.441.8+11.6
- SciCodeClaude Opus 4.6 by 951.942.9+9
- SWE-bench VerifiedClaude Opus 4.6 by 5.980.874.9+5.9
- LiveCodeBench v6Claude Opus 4.6 by 1.888.887+1.8
AgentsClaude Opus 4.6
- GDPValClaude Opus 4.6 by 34.955.921+34.9
- BrowseCompClaude Opus 4.6 by 29.18454.9+29.1
ReasoningClaude Opus 4.6
- ARC-AGI-2Claude Opus 4.6 by 58.968.89.9+58.9
- Humanity's Last ExamClaude Opus 4.6 by 11.439.928.5+11.4
- SimpleBenchClaude Opus 4.6 by 10.967.656.7+10.9
- CritPtClaude Opus 4.6 by 6.912.65.7+6.9
- GPQA DiamondClaude Opus 4.6 by 4.289.685.4+4.2
FactsClaude Opus 4.6
- AA-Omniscience · Non-hallucinationClaude Opus 4.6 by 19.437.217.8+19.4
- AA-Omniscience · AccuracyClaude Opus 4.6 by 6.74740.3+6.7
- SimpleQA VerifiedGPT-5 by 3.14750.1+3.1
- Vectara HHEM hallucination ratelower is betterClaude Opus 4.6 by 2.912.215.1+2.9
Images and chartsEven
- LMArena · VisionClaude Opus 4.6 by 108 rating points13161208+108 rating
- OCRBenchv2GPT-5 by 7.148.455.5+7.1
- CharXiv (RQ)GPT-5 by 3.777.481.1+3.7
- MMMU-ProClaude Opus 4.6 by 1.275.474.2+1.2
MathClaude Opus 4.6
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.6 by 10.66655.4+10.6
- FrontierMath Tier 4Claude Opus 4.6 by 4.826.821.9+4.8
- IMOAnswerBenchtie75.376tie
Following instructionsGPT-5
- Multi-ChallengeGPT-5 by 32.437.169.6+32.4
- IFBenchGPT-5 by 2053.173.1+20
Other results20 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 41.167.626.5+41.1
- Terminal-Bench 2.0Claude Opus 4.6 by 30.265.435.2+30.2
- ARC-AGI-1Claude Opus 4.6 by 27.39365.7+27.3
- ERQAGPT-5 by 24.940.865.7+24.9
- AA-OmniscienceClaude Opus 4.6 by 22.413.7-8.7+22.4
- HLE (with tools)Claude Opus 4.6 by 18.153.335.2+18.1
- τ²-Bench (Retail)Claude Opus 4.6 by 10.891.981.1+10.8
- frontiermath_tier_4_v1Claude Opus 4.6 by 10.422.912.5+10.4
- Artificial Analysis Coding IndexClaude Opus 4.6 by 10.348.137.8+10.3
- AA IntelligenceClaude Opus 4.6 by 8.931.923+8.9
- τ²-Bench Telecom (AA run)Claude Opus 4.6 by 7.392.184.8+7.3
- HMMT Nov. 2025Claude Opus 4.6 by 7.196.389.2+7.1
- matharena_visual_math_overallGPT-5 by 6.572.378.8+6.5
- LiveCodeBenchClaude Opus 4.6 by 5.388.883.5+5.3
- AIME 2025Claude Opus 4.6 by 5.299.894.6+5.2
- vectara_factual_consistencyClaude Opus 4.6 by 2.987.884.9+2.9
- vectara_avg_summary_lengthGPT-5 by 25.1 rating points137.6162.7+25.1 rating
- MMLU-ProClaude Opus 4.6 by 289.187.1+2
- AIME25 no toolsClaude Opus 4.6 by 195.694.6+1
- vectara_answer_ratetie99.899.9tie
Questions people ask
Which is better, Claude Opus 4.6 or GPT-5?
Claude Opus 4.6 wins five of the eight areas we test: coding, agents, reasoning, facts and math. GPT-5 wins following instructions, and costs 63% less. They are level on images and charts and long documents.
Which is better for coding?
Claude Opus 4.6. It wins 7 of the 7 coding tests both models report; GPT-5 wins none.
Which is cheaper?
Claude Opus 4.6 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5 costs $1.25 and $10.00. That makes GPT-5 about 63% cheaper for the same work.
How do you compare the two?
We use the 48 benchmark tests both models have published scores on. The verdict counts the 28 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 20 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.