Model comparison
GPT-5.4 vs GPT-5.4 mini
GPT-5.4 is the stronger model overall, scoring 59.5 to 44.8 on the Noometry Index. GPT-5.4 mini costs 3.3× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 45 shared benchmarks.
Summary
- They share 45 benchmarks with published results for both. GPT-5.4 scores higher in 10 categories and GPT-5.4 mini in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.7 to 30.3.
- The biggest single-benchmark swing is ARC-AGI-2: 74% for GPT-5.4 and 18.9% for GPT-5.4 mini.
- GPT-5.4 mini is cheaper at $0.75 / $4.50 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-5.4 | GPT-5.4 mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.5 | 44.8 |
| Released | 2026-03-05 | 2026-03-17 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 400K |
| Max output | 128K | 128K |
| Input $ / M tokens | $2.50 | $0.75 |
| Output $ / M tokens | $15 | $4.50 |
| Results tracked | 68 | 46 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5.4: 52.8 (#28), GPT-5.4 mini: 45.0 (#67)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena WebDev | 1465 | 1397 |
| SciCode | 56.6% | 49.9% |
| WeirdML | 77.7% | 60.3% |
| LMArena Coding | 1497 | 1438 |
| ALE-Bench | 1,607 | 1,189 |
| SWE-bench Verified | 76.9% | — |
| DeepSWE | 51.8% | — |
| FrontierCode | — | 27% |
| GSO | 31.4% | — |
| MirrorCode | 15.6% | — |
| AlgoTune | 1.85 | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#12), GPT-5.4 mini: 29.6 (#74)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| DeepResearch Bench | 35.1% | 36.3% |
| Terminal-Bench | 81.8% | — |
| APEX-Agents | 52.4% | — |
| τ²-bench Banking | 39.4% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| LMArena Search | 1197 | — |
| METR Time Horizons | 74.3% | — |
| Vending-Bench 2 | 6,144 | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.7 (#17), GPT-5.4 mini: 30.3 (#77)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| ARC-AGI-2 | 74% | 18.9% |
| Kagi LLM Benchmark | 63.8% | 37.9% |
| NYT Connections (extended) | 91.3% | 61.8% |
| ARC-AGI-1 | 93.7% | 63.7% |
| CritPt | 23.4% | 10% |
| Chess Puzzles | 44% | 24% |
| Thematic Generalization | 80% | 61.7% |
| LMArena Hard Prompts | 1485 | 1424 |
| Mystery Game Puzzles | 37% | 11% |
| DTBench | 94.4% | 80% |
| LMCA | 52% | 40.8% |
| Epoch Capabilities Index | 156.92 | 148.83 |
| ForecastBench | 59.5 | 57 |
| EnigmaEval | 16% | — |
| EBR-Bench | 25.4% | — |
Math GPT-5.4 leads
GPT-5.4: 73.3 (#14), GPT-5.4 mini: 45.4 (#63)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 51.2% |
| FrontierMath Tier 4 | 49% | 9.8% |
| OTIS Mock AIME 2024-2025 | 97.8% | 88.9% |
| ProofBench | 56% | 21% |
| LMArena Math | 1488 | 1418 |
| FrontierMath (Feb 2025 set) | 47.6% | 28.3% |
| FrontierMath Tier 4 (v1) | 27.1% | 2.1% |
| MathArena Final-Answer Competitions | 83.1% | — |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#13), GPT-5.4 mini: 51.4 (#62)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 93.3% | 86.9% |
| SimpleQA Verified | 45.1% | 29.4% |
| Vectara Hallucination Rate | 7% | 5.5% |
| LMArena Expert | 1507 | 1434 |
| Humanity's Last Exam | 36.2% | — |
Multimodal GPT-5.4 leads
GPT-5.4: 43.8 (#19), GPT-5.4 mini: 39.0 (#51)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena Vision | 1302 | 1245 |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.4 (#22), GPT-5.4 mini: 52.1 (#94)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena Non-English | 1465 | 1405 |
| LMArena Chinese | 1520 | 1446 |
| LMArena French | 1493 | 1440 |
| LMArena German | 1472 | 1407 |
| LMArena Japanese | 1484 | 1375 |
| LMArena Korean | 1449 | 1369 |
| LMArena Russian | 1480 | 1417 |
| LMArena Spanish | 1453 | 1403 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.3 (#25), GPT-5.4 mini: 74.3 (#99)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena Instruction Following | 1469 | 1405 |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), GPT-5.4 mini: 42.9 (#107)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena Longer Query | 1474 | 1407 |
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
Writing & Preference GPT-5.4 leads
GPT-5.4: 72.1 (#18), GPT-5.4 mini: 64.2 (#56)
| Benchmark | GPT-5.4 | GPT-5.4 mini |
|---|---|---|
| LMArena Text | 1469 | 1411 |
| LMArena Creative Writing | 1439 | 1370 |
| EQ-Bench Creative Writing | 1840 | 1665 |
| LMArena Multi-Turn | 1482 | 1429 |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than GPT-5.4 mini?
GPT-5.4 is the stronger model overall, scoring 59.5 to 44.8 on the Noometry Index. GPT-5.4 mini costs 3.3× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 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-5.4 lists at $2.50 and $15.
Is GPT-5.4 or GPT-5.4 mini better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.8 versus 45.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5.4 and GPT-5.4 mini share?
45 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and GPT-5.4 mini has 46.