Model comparison
GPT-5.4 mini vs o3
o3 is the stronger model overall, scoring 47.5 to 45.0 on the Noometry Index. GPT-5.4 mini costs 2.1× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Last verified . 38 shared benchmarks.
Summary
- They share 38 benchmarks with published results for both. GPT-5.4 mini scores higher in 3 categories and o3 in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where o3 leads 53.3 to 43.0.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 37.9% for GPT-5.4 mini and 67.6% for o3.
- GPT-5.4 mini is cheaper at $0.75 / $4.50 per million input/output tokens, against $2 / $8 for o3.
- GPT-5.4 mini accepts more context: 400K tokens versus 200K.
Side by side
| GPT-5.4 mini | o3 | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 45.0 | 47.5 |
| Released | 2026-03-17 | 2025-04-16 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 200K |
| Max output | 128K | 100K |
| Input $ / M tokens | $0.75 | $2 |
| Output $ / M tokens | $4.50 | $8 |
| Results tracked | 46 | 63 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3 leads
GPT-5.4 mini: 45.2 (#72), o3: 46.8 (#64)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| WeirdML | 60.3% | 52.4% |
| LMArena Coding | 1438 | 1408 |
| ALE-Bench | 1,189 | 933.55 |
| SWE-bench Verified | — | 62.3% |
| FrontierCode | 27% | — |
| SWE-bench Verified (bash only) | — | 58.4% |
| Aider Polyglot | — | 81.3% |
| LMArena WebDev | 1397 | — |
| SciCode | 49.9% | — |
| GSO | — | 8.8% |
| CadEval | — | 74% |
Agentic & Tool Use o3 leads
GPT-5.4 mini: 29.9 (#81), o3: 34.5 (#44)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| DeepResearch Bench | 36.3% | 45.2% |
| Berkeley Function Calling Leaderboard | — | 63% |
| GDPval | — | 30.8% |
| OSWorld | — | 23% |
| LMArena Search | — | 1144 |
| METR Time Horizons | — | 65.4% |
Reasoning o3 leads
GPT-5.4 mini: 30.4 (#85), o3: 32.0 (#78)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| ARC-AGI-2 | 18.9% | 6.5% |
| Kagi LLM Benchmark | 37.9% | 67.6% |
| ARC-AGI-1 | 63.7% | 60.8% |
| CritPt | 10% | 1.4% |
| Chess Puzzles | 24% | 38% |
| LMArena Hard Prompts | 1424 | 1402 |
| Mystery Game Puzzles | 11% | 29% |
| DTBench | 80% | 84.8% |
| LMCA | 40.8% | 39.7% |
| Epoch Capabilities Index | 148.84 | 146.86 |
| ForecastBench | 57 | 62.5 |
| SimpleBench | — | 53.1% |
| NYT Connections (extended) | 61.8% | — |
| EnigmaEval | — | 13.1% |
| Thematic Generalization | 61.7% | — |
Math o3 leads
GPT-5.4 mini: 45.5 (#75), o3: 50.2 (#58)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | 33.3% |
| OTIS Mock AIME 2024-2025 | 88.9% | 84.4% |
| LMArena Math | 1419 | 1426 |
| FrontierMath (Feb 2025 set) | 28.3% | 18.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 2.1% |
| FrontierMath Tier 4 | 9.8% | — |
| ProofBench | 21% | — |
| Omni-MATH | — | 71.4% |
| MATH Level 5 | — | 97.8% |
Knowledge o3 leads
GPT-5.4 mini: 51.5 (#67), o3: 54.6 (#52)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| GPQA Diamond | 86.9% | 81.8% |
| SimpleQA Verified | 29.4% | 49.4% |
| LMArena Expert | 1435 | 1402 |
| Humanity's Last Exam | — | 20.3% |
| MMLU-Pro | — | 85.9% |
| Confabulations | — | 14.4% |
| Vectara Hallucination Rate | 5.5% | — |
| GPQA (HELM) | — | 75.3% |
Multimodal o3 leads
GPT-5.4 mini: 39.7 (#56), o3: 41.4 (#36)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| LMArena Vision | 1245 | 1214 |
| GeoBench | — | 74% |
| VPCT | — | 52% |
Multilingual Too close to call
GPT-5.4 mini: 51.9 (#96), o3: 51.7 (#105)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| LMArena Non-English | 1405 | 1401 |
| LMArena Chinese | 1446 | 1437 |
| LMArena French | 1440 | 1430 |
| LMArena German | 1409 | 1420 |
| LMArena Japanese | 1374 | 1403 |
| LMArena Korean | 1368 | 1370 |
| LMArena Russian | 1417 | 1406 |
| LMArena Spanish | 1405 | 1395 |
Instruction Following GPT-5.4 mini leads
GPT-5.4 mini: 74.1 (#102), o3: 72.8 (#127)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| LMArena Instruction Following | 1405 | 1368 |
| IFEval | — | 86.9% |
Long Context o3 leads
GPT-5.4 mini: 43.0 (#112), o3: 53.3 (#6)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| LMArena Longer Query | 1407 | 1372 |
| Fiction.LiveBench | — | 88.9% |
| CL-bench | — | 17.8% |
Writing & Preference Too close to call
GPT-5.4 mini: 64.0 (#58), o3: 63.5 (#64)
| Benchmark | GPT-5.4 mini | o3 |
|---|---|---|
| LMArena Text | 1412 | 1410 |
| LMArena Creative Writing | 1370 | 1359 |
| EQ-Bench Creative Writing | 1665 | 1676 |
| LMArena Multi-Turn | 1429 | 1405 |
| Short-Story Creative Writing | — | 83.9% |
| WildBench | — | 86.1% |
Frequently asked questions
Is GPT-5.4 mini better than o3?
o3 is the stronger model overall, scoring 47.5 to 45.0 on the Noometry Index. GPT-5.4 mini costs 2.1× less per token, which makes it the better buy when o3's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 mini or o3?
GPT-5.4 mini is cheaper. It lists at $0.75 per million input tokens and $4.50 per million output tokens; o3 lists at $2 and $8.
Is GPT-5.4 mini or o3 better for coding?
o3 scores higher on coding benchmarks: 46.8 versus 45.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 mini does, with 400K tokens against 200K.
How many benchmarks do GPT-5.4 mini and o3 share?
38 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and o3 has 63.