Model comparison
GPT-5.4 vs GPT-5 Mini
GPT-5.4 is the stronger model overall, scoring 59.4 to 41.8 on the Noometry Index. GPT-5 Mini costs 8.2× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 47 shared benchmarks.
Summary
- They share 47 benchmarks with published results for both. GPT-5.4 scores higher in 10 categories and GPT-5 Mini in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.8 to 23.9.
- The biggest single-benchmark swing is ARC-AGI-2: 74% for GPT-5.4 and 4.4% for GPT-5 Mini.
- GPT-5 Mini is cheaper at $0.25 / $2 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 Mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.4 | 41.8 |
| Released | 2026-03-05 | 2025-08-07 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 400K |
| Max output | 128K | 128K |
| Input $ / M tokens | $2.50 | $0.25 |
| Output $ / M tokens | $15 | $2 |
| Results tracked | 68 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 leads
GPT-5.4: 52.6 (#33), GPT-5 Mini: 40.1 (#146)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| SWE-bench Verified | 76.9% | 64.7% |
| SciCode | 56.6% | 39.2% |
| WeirdML | 77.7% | 52.7% |
| LMArena Coding | 1497 | 1406 |
| ALE-Bench | 1,607 | 799.77 |
| AlgoTune | 1.85 | 1.38 |
| DeepSWE | 51.8% | — |
| SWE-bench Verified (bash only) | — | 59.8% |
| LMArena WebDev | 1465 | — |
| SWE-bench Multilingual | — | 39.7% |
| GSO | 31.4% | — |
| MirrorCode | 15.6% | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#13), GPT-5 Mini: 31.1 (#70)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| Terminal-Bench | 81.8% | 34.8% |
| Vending-Bench 2 | 6,144 | -31.18 |
| APEX-Agents | 52.4% | — |
| Berkeley Function Calling Leaderboard | — | 55.5% |
| τ²-bench Banking | 39.4% | — |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| LMArena Search | 1197 | — |
| METR Time Horizons | 74.3% | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.8 (#19), GPT-5 Mini: 23.9 (#168)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| ARC-AGI-2 | 74% | 4.4% |
| Kagi LLM Benchmark | 63.8% | 70.3% |
| ARC-AGI-1 | 93.7% | 54.3% |
| CritPt | 23.4% | 0% |
| Chess Puzzles | 44% | 30% |
| EnigmaEval | 16% | 8.2% |
| LMArena Hard Prompts | 1485 | 1380 |
| Mystery Game Puzzles | 37% | 10% |
| DTBench | 94.4% | 80.5% |
| LMCA | 52% | 34.2% |
| Epoch Capabilities Index | 156.81 | 145.52 |
| ForecastBench | 59.5 | 61 |
| NYT Connections (extended) | 91.3% | — |
| Thematic Generalization | 80% | — |
| EBR-Bench | 25.4% | — |
Math GPT-5.4 leads
GPT-5.4: 73.5 (#19), GPT-5 Mini: 46.7 (#69)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| FrontierMath (Tiers 1-3) | 78.6% | 46.7% |
| FrontierMath Tier 4 | 49% | 12.2% |
| OTIS Mock AIME 2024-2025 | 97.8% | 86.7% |
| ProofBench | 56% | 9% |
| LMArena Math | 1488 | 1378 |
| FrontierMath (Feb 2025 set) | 47.6% | 27.2% |
| FrontierMath Tier 4 (v1) | 27.1% | 6.3% |
| MathArena Final-Answer Competitions | 83.1% | — |
| Omni-MATH | — | 72.2% |
| MATH Level 5 | — | 97.8% |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#14), GPT-5 Mini: 45.6 (#86)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| GPQA Diamond | 93.3% | 75% |
| Humanity's Last Exam | 36.2% | 19.4% |
| SimpleQA Verified | 45.1% | 21.6% |
| Vectara Hallucination Rate | 7% | 12.9% |
| LMArena Expert | 1507 | 1379 |
| MMLU-Pro | — | 83.5% |
| Confabulations | — | 13.3% |
| GPQA (HELM) | — | 75.6% |
Multimodal GPT-5.4 leads
GPT-5.4: 43.7 (#20), GPT-5 Mini: 35.6 (#85)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| LMArena Vision | 1303 | 1202 |
| VPCT | — | 40.2% |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.2 (#23), GPT-5 Mini: 48.9 (#137)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| LMArena Non-English | 1465 | 1363 |
| LMArena Chinese | 1519 | 1385 |
| LMArena French | 1493 | 1386 |
| LMArena German | 1472 | 1366 |
| LMArena Japanese | 1485 | 1341 |
| LMArena Korean | 1448 | 1308 |
| LMArena Russian | 1480 | 1362 |
| LMArena Spanish | 1454 | 1355 |
Instruction Following Too close to call
GPT-5.4: 77.1 (#27), GPT-5 Mini: 76.2 (#46)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| LMArena Instruction Following | 1469 | 1357 |
| IFEval | — | 92.7% |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), GPT-5 Mini: 41.9 (#132)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| LMArena Longer Query | 1473 | 1355 |
| Fiction.LiveBench | — | 69.4% |
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
Writing & Preference GPT-5.4 leads
GPT-5.4: 71.9 (#17), GPT-5 Mini: 55.2 (#148)
| Benchmark | GPT-5.4 | GPT-5 Mini |
|---|---|---|
| LMArena Text | 1469 | 1373 |
| LMArena Creative Writing | 1439 | 1325 |
| EQ-Bench Creative Writing | 1840 | 1313 |
| LMArena Multi-Turn | 1482 | 1363 |
| Short-Story Creative Writing | — | 83.1% |
| WildBench | — | 85.5% |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than GPT-5 Mini?
GPT-5.4 is the stronger model overall, scoring 59.4 to 41.8 on the Noometry Index. GPT-5 Mini costs 8.2× 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 Mini?
GPT-5 Mini is cheaper. It lists at $0.25 per million input tokens and $2 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is GPT-5.4 or GPT-5 Mini better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.6 versus 40.1 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 Mini share?
47 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and GPT-5 Mini has 60.