GPT-5 vs GPT-5.4
Wins 0 of 8 areas
—
Wins 6 of 8 areas
Coding · Agents · Reasoning · Images and charts · Math · Long documents
GPT-5.4 is the stronger all-rounder.GPT-5 is cheaper.
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 software06GPT-5.46 of 6 tests
- ReasoningHard problems that need careful thinking04GPT-5.44 of 4 tests
- AgentsCarrying out multi-step tasks on its own03GPT-5.43 of 3 tests
- Images and chartsUnderstanding pictures, charts and video03GPT-5.43 of 3 tests
- MathCompetition and research-level math03GPT-5.43 of 3 tests
- Long documentsFinding answers in very long texts01GPT-5.41 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
- Following instructionsDoing exactly what it is asked00Even0 each · 1 tie
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 costs 36% 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
- Abstract visual puzzles that people can solveARC-AGI-2+63.4points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+43.1points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+27.8points ahead
Where GPT-5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+9.5points ahead
- Short factual questions, answered correctlySimpleQA Verified+5points 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.
CodingGPT-5.4
- Terminal-Bench 2.1GPT-5.4 by 43.135.278.3+43.1
- Terminal-Bench HardGPT-5.4 by 2532.657.6+25
- SWE-bench MultilingualGPT-5.4 by 16.455.371.7+16.4
- SWE-bench ProGPT-5.4 by 15.941.857.7+15.9
- SciCodeGPT-5.4 by 13.742.956.6+13.7
- LMArena · WebDevGPT-5.4 by 47 rating points14181465+47 rating
AgentsGPT-5.4
- BrowseCompGPT-5.4 by 27.854.982.7+27.8
- τ-Bench V3 · BankingGPT-5.4 by 17.522.139.6+17.5
- GDPValGPT-5.4 by 16.42137.4+16.4
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 63.49.973.3+63.4
- CritPtGPT-5.4 by 17.75.723.4+17.7
- Humanity's Last ExamGPT-5.4 by 15.228.543.7+15.2
- GPQA DiamondGPT-5.4 by 6.685.492+6.6
FactsEven
- AA-Omniscience · AccuracyGPT-5.4 by 10.640.350.9+10.6
- AA-Omniscience · Non-hallucinationGPT-5 by 9.517.88.3+9.5
- Vectara HHEM hallucination ratelower is betterGPT-5.4 by 8.115.17+8.1
- SimpleQA VerifiedGPT-5 by 550.145.1+5
Images and chartsGPT-5.4
- LMArena · VisionGPT-5.4 by 94 rating points12081302+94 rating
- MMMU-ProGPT-5.4 by 4.274.278.4+4.2
- CharXiv (RQ)GPT-5.4 by 1.781.182.8+1.7
MathGPT-5.4
- FrontierMath Tier 4GPT-5.4 by 2721.949+27
- FrontierMath Tiers 1-3 (v2)GPT-5.4 by 23.255.478.6+23.2
- IMOAnswerBenchGPT-5.4 by 15.47691.4+15.4
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.
- Terminal-Bench 2.0GPT-5.4 by 39.935.275.1+39.9
- Artificial Analysis Coding IndexGPT-5.4 by 33.337.871.1+33.3
- ARC-AGI-1GPT-5.4 by 2865.793.7+28
- FrontierMath (overall)GPT-5.4 by 21.326.347.6+21.3
- AA Agentic IndexGPT-5.4 by 17.726.544.2+17.7
- HLE (with tools)GPT-5.4 by 16.835.252+16.8
- AA IntelligenceGPT-5.4 by 162339+16
- frontiermath_tier_4_v1GPT-5.4 by 14.612.527.1+14.6
- AA-OmniscienceGPT-5.4 by 14.5-8.75.8+14.5
- matharena_visual_math_overallGPT-5.4 by 13.778.892.5+13.7
- vectara_avg_summary_lengthGPT-5 by 81 rating points162.781.7+81 rating
- vectara_factual_consistencyGPT-5.4 by 8.184.993+8.1
- HMMT Nov. 2025GPT-5.4 by 6.689.295.8+6.6
- HealthBench Hard length-adjustedGPT-5 by 5.634.729.1+5.6
- HealthBench length-adjustedGPT-5 by 3.757.754+3.7
- τ²-Bench Telecom (AA run)GPT-5.4 by 2.384.887.1+2.3
- HealthBench Professional length-adjustedGPT-5.4 by 1.946.248.1+1.9
- HealthBench Consensus length-adjustedtie95.696.3tie
- MMLU-Protie87.187.5tie
- vectara_answer_ratetie99.999.9tie
Questions people ask
Which is better, GPT-5 or GPT-5.4?
GPT-5.4 wins six of the eight areas we test: coding, agents, reasoning, images and charts, math and long documents. GPT-5 wins none, but costs 36% less. They are level on facts and following instructions.
Which is better for coding?
GPT-5.4. It wins 6 of the 6 coding tests both models report; GPT-5 wins none.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; GPT-5.4 costs $2.50 and $15.00. That makes GPT-5 about 36% 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 25 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.