GPT-5.4 mini vs GPT-5.4 nano
Wins 4 of 8 areas
Coding · Agents · Reasoning · Images and charts
Wins 2 of 8 areas
Math · Following instructions
GPT-5.4 mini wins more areas, narrowly.GPT-5.4 nano is cheaper and better at following instructions.
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.
- AgentsCarrying out multi-step tasks on its own61GPT-5.4 mini6 of 7 tests
- Images and chartsUnderstanding pictures, charts and video20GPT-5.4 mini2 of 2 tests
- CodingWriting and fixing software32GPT-5.4 mini3 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking21GPT-5.4 mini2 of 5 tests · 2 ties
- Following instructionsDoing exactly what it is asked02GPT-5.4 nano2 of 2 tests
- MathCompetition and research-level math12GPT-5.4 nano2 of 3 tests
- FactsGetting facts right instead of making them up22Even2 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-5.4 nano costs 72% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.4 mini pulls ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+33.1points ahead
- Short factual questions, answered correctlySimpleQA Verified+17.7points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+13.2points ahead
Where GPT-5.4 nano pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+16points ahead
- Fresh competition math problemsLiveBench · Mathematics+12.5points ahead
- Fresh logic puzzles that change regularlyLiveBench · Reasoning+9.8points 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 mini
- Terminal-Bench HardGPT-5.4 mini by 9.952.342.4+9.9
- LiveBench · Agentic CodingGPT-5.4 nano by 5.141.746.8+5.1
- SciCodeGPT-5.4 mini by 4.952.147.2+4.9
- SWE-bench ProGPT-5.4 mini by 254.452.4+2
- Terminal-Bench 2.1GPT-5.4 nano by 1.559.260.7+1.5
- LiveBench · Codingtie71.670.8tie
AgentsGPT-5.4 mini
- OSWorld-VerifiedGPT-5.4 mini by 33.172.139+33.1
- AA IT-Bench SREGPT-5.4 mini by 10.835.224.4+10.8
- AA ApexAgentsGPT-5.4 mini by 3.328.224.9+3.3
- GDPValGPT-5.4 mini by 3.325.822.5+3.3
- τ-Bench V3 · BankingGPT-5.4 nano by 1.825.627.4+1.8
- MCP AtlasGPT-5.4 mini by 1.657.756.1+1.6
- Terminal-Bench 4.0GPT-5.4 mini by 1.520.5+1.5
ReasoningGPT-5.4 mini
- ARC-AGI-2GPT-5.4 mini by 13.218.95.7+13.2
- LiveBench · ReasoningGPT-5.4 nano by 9.871.381.1+9.8
- GPQA DiamondGPT-5.4 mini by 5.887.581.7+5.8
- CritPttie109.3tie
- Humanity's Last Examtie28.128.3tie
FactsEven
- SimpleQA VerifiedGPT-5.4 mini by 17.729.411.7+17.7
- AA-Omniscience · Non-hallucinationGPT-5.4 nano by 169.825.8+16
- AA-Omniscience · AccuracyGPT-5.4 mini by 11.837.525.7+11.8
- Vectara HHEM hallucination ratelower is betterGPT-5.4 nano by 2.45.53.1+2.4
Images and chartsGPT-5.4 mini
- MMMU-ProGPT-5.4 mini by 7.973.365.4+7.9
- LMArena · VisionGPT-5.4 mini by 49 rating points12451196+49 rating
MathGPT-5.4 nano
- LiveBench · MathematicsGPT-5.4 nano by 12.578.591+12.5
- FrontierMath Tiers 1-3 (v2)GPT-5.4 mini by 6.351.244.9+6.3
- FrontierMath Tier 4GPT-5.4 nano by 2.49.812.2+2.4
Following instructionsGPT-5.4 nano
- LiveBench · Instruction FollowingGPT-5.4 nano by 7.459.867.2+7.4
- IFBenchGPT-5.4 nano by 2.673.375.9+2.6
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.
- Terminal-Bench 2.0GPT-5.4 mini by 13.76046.3+13.7
- ARC-AGI-1GPT-5.4 mini by 12.263.751.5+12.2
- OmniDocBench 1.5GPT-5.4 mini by 11.687.475.8+11.6
- AA-OmniscienceGPT-5.4 mini by 10.6-18.9-29.5+10.6
- vectara_avg_summary_lengthGPT-5.4 nano by 89.7 rating points54.7144.4+89.7 rating
- livebench_languageGPT-5.4 mini by 8.57162.5+8.5
- ToolathlonGPT-5.4 mini by 7.442.935.5+7.4
- τ²-Bench Telecom (AA run)GPT-5.4 mini by 7.383.376+7.3
- Finance Agent v2GPT-5.4 mini by 7.245.438.2+7.2
- frontiermath_tier_4_v1GPT-5.4 nano by 4.22.16.3+4.2
- AA IntelligenceGPT-5.4 mini by 3.424.120.7+3.4
- livebench_data_analysisGPT-5.4 mini by 3.270.867.6+3.2
- vectara_factual_consistencyGPT-5.4 nano by 2.494.596.9+2.4
- AA Agentic IndexGPT-5.4 mini by 219.717.7+2
- Artificial Analysis Coding Indextie56.156.1tie
- vectara_answer_ratetie100100tie
Questions people ask
Which is better, GPT-5.4 mini or GPT-5.4 nano?
GPT-5.4 mini wins four of the eight areas we test: coding, agents, reasoning and images and charts. GPT-5.4 nano wins math and following instructions, and costs 72% less. They are level on facts and long documents.
Which is better for coding?
GPT-5.4 mini. It wins 3 of the 6 coding tests both models report; GPT-5.4 nano wins 2, and 1 is a tie.
Which is cheaper?
GPT-5.4 mini costs $0.75 per million input tokens and $4.50 per million output tokens; GPT-5.4 nano costs $0.20 and $1.25. That makes GPT-5.4 nano about 72% 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.