Model comparison
GPT-5 vs GPT-5.4 mini
GPT-5 is the stronger model overall, scoring 50.9 to 44.8 on the Noometry Index. GPT-5.4 mini costs 2.0× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Last verified . 43 shared benchmarks.
Summary
- They share 43 benchmarks with published results for both. GPT-5 scores higher in 7 categories and GPT-5.4 mini in 3 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 42.9.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 72.7% for GPT-5 and 37.9% for GPT-5.4 mini.
- GPT-5.4 mini is cheaper at $0.75 / $4.50 per million input/output tokens, against $1.25 / $10 for GPT-5.
Side by side
| GPT-5 | GPT-5.4 mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 50.9 | 44.8 |
| Released | 2025-08-07 | 2026-03-17 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 400K |
| Max output | 128K | 128K |
| Input $ / M tokens | $1.25 | $0.75 |
| Output $ / M tokens | $10 | $4.50 |
| Results tracked | 69 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5 leads
GPT-5: 50.7 (#39), GPT-5.4 mini: 45.0 (#67)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena WebDev | 1417 | 1397 |
| SciCode | 42.9% | 49.9% |
| WeirdML | 60.7% | 60.3% |
| LMArena Coding | 1436 | 1438 |
| ALE-Bench | 1,162 | 1,189 |
| SWE-bench Verified | 73.6% | — |
| FrontierCode | — | 27% |
| SWE-bench Verified (bash only) | 65% | — |
| Aider Polyglot | 88% | — |
| GSO | 6.9% | — |
| AlgoTune | 1.67 | — |
Agentic & Tool Use GPT-5 leads
GPT-5: 33.0 (#53), GPT-5.4 mini: 29.6 (#74)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| DeepResearch Bench | 49.6% | 36.3% |
| Terminal-Bench | 49.6% | — |
| GDPval | 34.8% | — |
| Remote Labor Index | 1.7% | — |
| BALROG | 32.8% | — |
| LMArena Search | 1133 | — |
| METR Time Horizons | 69.6% | — |
Reasoning GPT-5 leads
GPT-5: 38.3 (#55), GPT-5.4 mini: 30.3 (#77)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 9.9% | 18.9% |
| Kagi LLM Benchmark | 72.7% | 37.9% |
| ARC-AGI-1 | 65.7% | 63.7% |
| CritPt | 12.6% | 10% |
| Chess Puzzles | 37% | 24% |
| LMArena Hard Prompts | 1415 | 1424 |
| Mystery Game Puzzles | 23% | 11% |
| DTBench | 90.7% | 80% |
| LMCA | 40% | 40.8% |
| Epoch Capabilities Index | 150 | 148.83 |
| ForecastBench | 61.4 | 57 |
| SimpleBench | 56.7% | — |
| NYT Connections (extended) | — | 61.8% |
| EnigmaEval | 10.5% | — |
| Thematic Generalization | — | 61.7% |
| EBR-Bench | 12.7% | — |
Math GPT-5 leads
GPT-5: 54.6 (#36), GPT-5.4 mini: 45.4 (#63)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 55.4% | 51.2% |
| FrontierMath Tier 4 | 22% | 9.8% |
| OTIS Mock AIME 2024-2025 | 91.4% | 88.9% |
| ProofBench | 18% | 21% |
| LMArena Math | 1406 | 1418 |
| FrontierMath (Feb 2025 set) | 32.4% | 28.3% |
| FrontierMath Tier 4 (v1) | 12.5% | 2.1% |
| Omni-MATH | 64.7% | — |
| MATH Level 5 | 98.1% | — |
Knowledge GPT-5 leads
GPT-5: 56.5 (#37), GPT-5.4 mini: 51.4 (#62)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 86.2% | 86.9% |
| SimpleQA Verified | 50.1% | 29.4% |
| Vectara Hallucination Rate | 14.7% | 5.5% |
| LMArena Expert | 1420 | 1434 |
| Humanity's Last Exam | 25.3% | — |
| MMLU-Pro | 86.3% | — |
| Confabulations | 10.3% | — |
| GPQA (HELM) | 79.2% | — |
Multimodal GPT-5 leads
GPT-5: 46.1 (#12), GPT-5.4 mini: 39.0 (#51)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | 1232 | 1245 |
| GeoBench | 81% | — |
| VPCT | 66% | — |
Multilingual Too close to call
GPT-5: 51.5 (#105), GPT-5.4 mini: 52.1 (#94)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1397 | 1405 |
| LMArena Chinese | 1423 | 1446 |
| LMArena French | 1409 | 1440 |
| LMArena German | 1415 | 1407 |
| LMArena Japanese | 1411 | 1375 |
| LMArena Korean | 1361 | 1369 |
| LMArena Russian | 1406 | 1417 |
| LMArena Spanish | 1398 | 1403 |
Instruction Following Too close to call
GPT-5: 74.0 (#107), GPT-5.4 mini: 74.3 (#99)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1388 | 1405 |
| IFEval | 87.5% | — |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), GPT-5.4 mini: 42.9 (#107)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1399 | 1407 |
| Fiction.LiveBench | 97.2% | — |
Writing & Preference Too close to call
GPT-5: 63.5 (#63), GPT-5.4 mini: 64.2 (#56)
| Benchmark | GPT-5 | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1406 | 1411 |
| LMArena Creative Writing | 1365 | 1370 |
| EQ-Bench Creative Writing | 1627 | 1665 |
| LMArena Multi-Turn | 1426 | 1429 |
| Short-Story Creative Writing | 86% | — |
| WildBench | 85.7% | — |
Frequently asked questions
Is GPT-5 better than GPT-5.4 mini?
GPT-5 is the stronger model overall, scoring 50.9 to 44.8 on the Noometry Index. GPT-5.4 mini costs 2.0× less per token, which makes it the better buy when GPT-5's lead doesn't matter for your workload.
Which is cheaper, GPT-5 or GPT-5.4 mini?
GPT-5.4 mini is cheaper. It lists at $0.75 per million input tokens and $4.50 per million output tokens; GPT-5 lists at $1.25 and $10.
Is GPT-5 or GPT-5.4 mini better for coding?
GPT-5 scores higher on coding benchmarks: 50.7 versus 45.0 in the Noometry coding category.
Which has the bigger context window?
Both accept 400K tokens.
How many benchmarks do GPT-5 and GPT-5.4 mini share?
43 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and GPT-5.4 mini has 46.