GPT-5.4 vs GPT-5.4 mini
Wins 7 of 8 areas
Coding · Agents · Reasoning · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
GPT-5.4 is the stronger all-rounder.GPT-5.4 mini is cheaper.
Scores updated · 46 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 software70GPT-5.47 of 7 tests
- AgentsCarrying out multi-step tasks on its own50GPT-5.45 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking50GPT-5.45 of 5 tests
- MathCompetition and research-level math30GPT-5.43 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20GPT-5.42 of 2 tests
- Following instructionsDoing exactly what it is asked10GPT-5.41 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10GPT-5.41 of 1 test
- FactsGetting facts right instead of making them up22Even2 each
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 mini costs 70% 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+54.4points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+39.2points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+27.4points ahead
Where GPT-5.4 mini pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+1.5points ahead
Every test, side by side
All 46 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 19.178.359.2+19.1
- LiveBench · Agentic CodingGPT-5.4 by 12.153.841.7+12.1
- LMArena · WebDevGPT-5.4 by 68 rating points14651397+68 rating
- LiveBench · CodingGPT-5.4 by 5.977.571.6+5.9
- Terminal-Bench HardGPT-5.4 by 5.357.652.3+5.3
- SciCodeGPT-5.4 by 4.556.652.1+4.5
- SWE-bench ProGPT-5.4 by 3.357.754.4+3.3
AgentsGPT-5.4
- τ-Bench V3 · BankingGPT-5.4 by 1439.625.6+14
- GDPValGPT-5.4 by 11.637.425.8+11.6
- MCP AtlasGPT-5.4 by 9.567.257.7+9.5
- AA ApexAgentsGPT-5.4 by 5.133.328.2+5.1
- OSWorld-VerifiedGPT-5.4 by 2.97572.1+2.9
- AA IT-Bench SREtie34.535.2tie
ReasoningGPT-5.4
- ARC-AGI-2GPT-5.4 by 54.473.318.9+54.4
- LiveBench · ReasoningGPT-5.4 by 16.888.171.3+16.8
- Humanity's Last ExamGPT-5.4 by 15.643.728.1+15.6
- CritPtGPT-5.4 by 13.423.410+13.4
- GPQA DiamondGPT-5.4 by 4.59287.5+4.5
FactsEven
- SimpleQA VerifiedGPT-5.4 by 15.745.129.4+15.7
- AA-Omniscience · AccuracyGPT-5.4 by 13.450.937.5+13.4
- AA-Omniscience · Non-hallucinationGPT-5.4 mini by 1.58.39.8+1.5
- Vectara HHEM hallucination ratelower is betterGPT-5.4 mini by 1.575.5+1.5
Images and chartsGPT-5.4
- LMArena · VisionGPT-5.4 by 57 rating points13021245+57 rating
- MMMU-ProGPT-5.4 by 5.178.473.3+5.1
MathGPT-5.4
- FrontierMath Tier 4GPT-5.4 by 39.2499.8+39.2
- FrontierMath Tiers 1-3 (v2)GPT-5.4 by 27.478.651.2+27.4
- LiveBench · MathematicsGPT-5.4 by 15.794.278.5+15.7
Following instructionsGPT-5.4
- LiveBench · Instruction FollowingGPT-5.4 by 10.470.259.8+10.4
- IFBenchtie73.973.3tie
Other results16 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 3093.763.7+30
- vectara_avg_summary_lengthGPT-5.4 by 2781.754.7+27
- frontiermath_tier_4_v1GPT-5.4 by 2527.12.1+25
- AA-OmniscienceGPT-5.4 by 24.75.8-18.9+24.7
- AA Agentic IndexGPT-5.4 by 24.544.219.7+24.5
- Terminal-Bench 2.0GPT-5.4 by 15.175.160+15.1
- Artificial Analysis Coding IndexGPT-5.4 by 1571.156.1+15
- AA IntelligenceGPT-5.4 by 14.93924.1+14.9
- ToolathlonGPT-5.4 by 11.754.642.9+11.7
- livebench_languageGPT-5.4 by 11.682.671+11.6
- livebench_data_analysisGPT-5.4 by 8.579.370.8+8.5
- τ²-Bench Telecom (AA run)GPT-5.4 by 3.887.183.3+3.8
- OmniDocBench 1.5GPT-5.4 by 1.789.187.4+1.7
- vectara_factual_consistencyGPT-5.4 mini by 1.59394.5+1.5
- Legal Agent Benchmarktie0.40tie
- vectara_answer_ratetie99.9100tie
Questions people ask
Which is better, GPT-5.4 or GPT-5.4 mini?
GPT-5.4 wins seven of the eight areas we test: coding, agents, reasoning, images and charts, math, long documents and following instructions. GPT-5.4 mini wins none, but costs 70% less. They are level on facts.
Which is better for coding?
GPT-5.4. It wins 7 of the 7 coding tests both models report; GPT-5.4 mini wins none.
Which is cheaper?
GPT-5.4 costs $2.50 per million input tokens and $15.00 per million output tokens; GPT-5.4 mini costs $0.75 and $4.50. That makes GPT-5.4 mini about 70% cheaper for the same work.
How do you compare the two?
We use the 46 benchmark tests both models have published scores on. The verdict counts the 30 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 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.