Claude 3.7 Sonnet vs GPT-5
Wins 1 of 7 areas
Agents
Wins 5 of 7 areas
Coding · Reasoning · Images and charts · Long documents · Following instructions
GPT-5 is the stronger all-rounder.Claude 3.7 Sonnet is better at agents.
Scores updated · 35 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.
- ReasoningHard problems that need careful thinking04GPT-54 of 4 tests
- CodingWriting and fixing software03GPT-53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03GPT-53 of 3 tests
- Following instructionsDoing exactly what it is asked02GPT-52 of 2 tests
- Long documentsFinding answers in very long texts01GPT-51 of 1 test
- AgentsCarrying out multi-step tasks on its own10Claude 3.7 Sonnet1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
Agents and long documents 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 costs 38% 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 GPT-5 pulls ahead
- Reasons across sets of long documentsAA-LCR+26.5points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+24.8points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+18.8points ahead
Where Claude 3.7 Sonnet pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+42.2points ahead
- Real work tasks from 44 professionsGDPVal+6.4points ahead
Every test, side by side
All 35 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5
- Terminal-Bench HardGPT-5 by 11.421.232.6+11.4
- SWE-bench VerifiedGPT-5 by 4.670.374.9+4.6
- SciCodeGPT-5 by 2.640.342.9+2.6
ReasoningGPT-5
- Humanity's Last ExamGPT-5 by 18.89.728.5+18.8
- SimpleBenchGPT-5 by 10.346.456.7+10.3
- GPQA DiamondGPT-5 by 8.277.285.4+8.2
- CritPtGPT-5 by 4.80.95.7+4.8
FactsEven
- AA-Omniscience · Non-hallucinationClaude 3.7 Sonnet by 42.26017.8+42.2
- AA-Omniscience · AccuracyGPT-5 by 12.32840.3+12.3
Images and chartsGPT-5
- MMMU-ProGPT-5 by 14.160.174.2+14.1
- MMMUGPT-5 by 9.27584.2+9.2
- LMArena · VisionGPT-5 by 39 rating points11691208+39 rating
Following instructionsGPT-5
- IFBenchGPT-5 by 24.848.373.1+24.8
- Multi-ChallengeGPT-5 by 1851.669.6+18
Other results19 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 Feb. 2025GPT-5 by 56.631.788.3+56.6
- metr_re_benchGPT-5 by 40040+40
- AIME 2025GPT-5 by 39.854.894.6+39.8
- ARC-AGI-1GPT-5 by 37.128.665.7+37.1
- τ²-Bench Telecom (AA run)GPT-5 by 30.154.784.8+30.1
- Aider-PolyglotGPT-5 by 27.660.488+27.6
- metr_hcastGPT-5 by 14.254.268.4+14.2
- HarmBenchGPT-5 by 13.384.397.6+13.3
- swe_bench_bashGPT-5 by 12.252.865+12.2
- AA Agentic IndexClaude 3.7 Sonnet by 10.53726.5+10.5
- Terminal-BenchGPT-5 by 8.635.243.8+8.6
- AA-OmniscienceClaude 3.7 Sonnet by 8-0.7-8.7+8
- AIR-Bench 2024GPT-5 by 5.981.887.7+5.9
- AA IntelligenceGPT-5 by 5.317.723+5.3
- bbqGPT-5 by 4.792.196.8+4.7
- Artificial Analysis Coding IndexGPT-5 by 1.436.437.8+1.4
- XSTesttie96.497.1tie
- anthropic_red_teamtie99.799.1tie
- simple_safety_teststie10099.8tie
Questions people ask
Which is better, Claude 3.7 Sonnet or GPT-5?
GPT-5 wins five of the seven areas where both have results: coding, reasoning, images and charts, long documents and following instructions. Claude 3.7 Sonnet wins agents. They are level on facts.
Which is better for coding?
GPT-5. It wins 3 of the 3 coding tests both models report; Claude 3.7 Sonnet wins none.
Which is cheaper?
Claude 3.7 Sonnet costs $3.00 per million input tokens and $15.00 per million output tokens; GPT-5 costs $1.25 and $10.00. That makes GPT-5 about 38% cheaper for the same work.
How do you compare the two?
We use the 35 benchmark tests both models have published scores on. The verdict counts the 16 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.