GPT-5.2 vs GPT-5.4
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | GPT-5.2 | GPT-5.4 |
|---|---|---|
| Coding | 26.0 GPT-5.2 | 25.9 GPT-5.4 |
| Agents | 59.5 GPT-5.2 | 62.0 GPT-5.4 |
| Reasoning | 55.3 GPT-5.2 | 61.4 GPT-5.4 |
| Knowledge & language | 38.1 GPT-5.2 | 46.9 GPT-5.4 |
| Human preference | 51.6 GPT-5.2 | 53.1 GPT-5.4 |
Benchmarks both report
GPT-5.4 leads on 25 of the 29 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | GPT-5.2 | GPT-5.4 |
|---|---|---|
| LMArena Elo | 1437 | 1475 |
| LMArena WebDev (LMArena) | 1419 | 1462 |
| MirrorCode (Epoch AI) | 32.2% | 23.3% |
| LMArena Vision (LMArena) | 1242 | 1282 |
| Chess Puzzles (Epoch AI) | 48% | 48% |
| EBR-bench (Epoch AI) | 23% | 25.4% |
| FrontierMath Tier 4 v2 (Epoch AI) | 31.7% | 49% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 67.4% | 78.6% |
| Furniture Assembly (Epoch AI) | 38.3% | 37.5% |
| GPQA Diamond | 92.4% | 92.8% |
| LiveBench Agentic Coding (LiveBench) | 50.3% | 53.8% |
| LiveBench Coding (LiveBench) | 76.1% | 77.5% |
| LiveBench Data Analysis (LiveBench) | 78.2% | 79.3% |
| LiveBench Instruction Following (LiveBench) | 61.8% | 70.2% |
| LiveBench Language (LiveBench) | 79.8% | 82.6% |
| LiveBench Mathematics (LiveBench) | 93.2% | 94.1% |
| LiveBench Reasoning (LiveBench) | 83.2% | 88.1% |
| Mystery Game Puzzles (Epoch AI) | 23% | 37% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 96.1% | 97.8% |
| SAGE (Vals AI) | 49.3% | 43.3% |
| SimpleQA Verified | 37.1% | 45.1% |
| SWE-bench Verified (Epoch AI) | 73.8% | 76.9% |
| Vending-Bench 2 (Andon Labs) | 3591.33 | 6144.18 |
| WeirdML (Håvard Tveit Ihle) | 72.2% | 77.7% |
| AIME 2026 | 98.3% | 99.2% |
| tau2-bench Banking Knowledge (Sierra) | 32.2% | 39.4% |
| GSO Opt@1 (GSO) | 26.5% | 30.4% |
| GDPval (wins or ties) | 70.9% | 83% |
| SWE-Bench Pro | 55.6% | 57.7% |
FAQ
Is GPT-5.2 better than GPT-5.4?
GPT-5.4 scores 50.0 on the Respan Index (#43), ahead of GPT-5.2 at 46.8 (#50). GPT-5.4 leads on 25 of the 29 benchmarks both report.
Which is cheaper, GPT-5.2 or GPT-5.4?
GPT-5.2 lists at $1.75 input and $14 output per 1M tokens; GPT-5.4 at $2.5 and $15. GPT-5.2 is 1.2x cheaper per token at list price.
Which has the larger context window?
GPT-5.2 accepts 400K tokens and GPT-5.4 1.1M tokens.