Claude Sonnet 4 vs GPT-5.4
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
GPT-5.4 is the stronger all-rounder.
Scores updated · 33 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 software05GPT-5.45 of 5 tests
- AgentsCarrying out multi-step tasks on its own04GPT-5.44 of 4 tests
- ReasoningHard problems that need careful thinking04GPT-5.44 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02GPT-5.42 of 2 tests
- FactsGetting facts right instead of making them up12GPT-5.42 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.41 of 1 test
- Following instructionsDoing exactly what it is asked01GPT-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 3% 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-5.4 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+70.5points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+67.4points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+42points ahead
Where Claude Sonnet 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+62.6points ahead
Every test, side by side
All 33 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.4
- Terminal-Bench 2.1GPT-5.4 by 4236.378.3+42
- Terminal-Bench HardGPT-5.4 by 26.531.157.6+26.5
- SciCodeGPT-5.4 by 16.64056.6+16.6
- SWE-bench ProGPT-5.4 by 1542.757.7+15
- SWE-bench MultilingualGPT-5.4 by 14.856.971.7+14.8
AgentsGPT-5.4
- BrowseCompGPT-5.4 by 70.512.282.7+70.5
- OSWorld-VerifiedGPT-5.4 by 32.842.275+32.8
- GDPValGPT-5.4 by 27.79.737.4+27.7
- τ-Bench V3 · BankingGPT-5.4 by 22.916.739.6+22.9
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 67.45.973.3+67.4
- Humanity's Last ExamGPT-5.4 by 3310.743.7+33
- CritPtGPT-5.4 by 23.10.323.4+23.1
- GPQA DiamondGPT-5.4 by 14.377.792+14.3
FactsGPT-5.4
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 62.670.98.3+62.6
- AA-Omniscience · AccuracyGPT-5.4 by 28.222.750.9+28.2
- Vectara HHEM hallucination ratelower is betterGPT-5.4 by 3.310.37+3.3
Images and chartsGPT-5.4
- MMMU-ProGPT-5.4 by 16.661.878.4+16.6
- LMArena · VisionGPT-5.4 by 111 rating points11911302+111 rating
Other results13 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.
- ARC-AGI-1GPT-5.4 by 53.74093.7+53.7
- Artificial Analysis Coding IndexGPT-5.4 by 33.537.671.1+33.5
- HLE (with tools)GPT-5.4 by 31.720.352+31.7
- frontiermath_tier_4_v1GPT-5.4 by 27.1027.1+27.1
- AA Agentic IndexGPT-5.4 by 26.617.644.2+26.6
- τ²-Bench Telecom (AA run)GPT-5.4 by 22.564.687.1+22.5
- AA IntelligenceGPT-5.4 by 20.118.939+20.1
- vectara_avg_summary_lengthClaude Sonnet 4 by 64.1 rating points145.881.7+64.1 rating
- AA-OmniscienceGPT-5.4 by 5.60.25.8+5.6
- LiveBenchGPT-5.4 by 5.574.880.3+5.5
- MMLU-ProGPT-5.4 by 3.883.787.5+3.8
- vectara_factual_consistencyGPT-5.4 by 3.389.793+3.3
- vectara_answer_rateGPT-5.4 by 1.398.699.9+1.3
Questions people ask
Which is better, Claude Sonnet 4 or GPT-5.4?
GPT-5.4 wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, long documents and following instructions. Claude Sonnet 4 wins none.
Which is better for coding?
GPT-5.4. It wins 5 of the 5 coding tests both models report; Claude Sonnet 4 wins none.
Which is cheaper?
Claude Sonnet 4 costs $3.00 per million input tokens and $15.00 per million output tokens; GPT-5.4 costs $2.50 and $15.00. That makes GPT-5.4 about 3% cheaper for the same work.
How do you compare the two?
We use the 33 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 13 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.