Up to four models
GPT-5.4 Nano vs GPT-5 nano
On the index (GPT-4 = 100), GPT-5.4 Nano scores 141 and GPT-5 nano 125. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GPT-5.4 Nano
141125–157
WeightsClosed
ReleasedMar 2026
Price in / out$0.20 / $1.25
Context400K
Core tests taken4 of 9
GPT-5 nano
125109–141
WeightsClosed
ReleasedAug 2025
Price in / out$0.050 / $0.40
Context400K
Core tests taken3 of 9
GPT-5.4 Nano leads GPT-5 nano by 17 points, ahead on 3 of 3 shared core tests, at 3.4× the price.
Test by testthird-party results
GPT-5.4 NanoGPT-5 nano
Every result as a table
| Test | GPT-5.4 Nano | GPT-5 nano |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 78.5% | 69.4% |
| GPQA Diamond (Vals AI) | 77.5% | 63.4% |
| MMLU Pro | 77.2% | 76.1% |
| MMMU Pro | 73.6% | 70.9% |
| SimpleQA Verified | 11.7% | 11.7% |
| Maths | ||
| FrontierMath (tiers 1-3) | 44.9% | 20.0% |
| AIME-style maths (OTIS mock) | 87.8% | 81.1% |
| Chess puzzles | 45.0% | 22.0% |
| FrontierMath (2025 set) | 25.9% | 8.3% |
| FrontierMath tier 4 (2025 set) | 6.3% | 2.1% |
| FrontierMath (tier 4) | 12.2% | 2.4% |
| Coding | ||
| LiveCodeBench | 84.0% | 70.2% |
| WeirdML | 49.2% | 38.1% |
| Agents | ||
| DTBench | 80.3% | 62.7% |
| LMCA | 36.9% | 7.9% |
| Mystery game puzzles | 9.0% | 9.0% |
| Professional work | ||
| LegalBench | 77.9% | 50.1% |
| MedCode | 41.0% | 30.4% |
| MedScribe | 77.1% | 72.9% |
| MortgageTax | 59.1% | 53.6% |
| TaxEval v2 | 67.4% | 67.4% |
| Novel problems | ||
| ARC-AGI-2 | 5.7% | 2.6% |
| ARC-AGI-1 | 51.5% | 20.7% |
Tests at least two of these models have taken. Bold rows are the core tests.