Model comparison
GPT-5.4 vs gpt-oss-120b
GPT-5.4 is the stronger model overall, scoring 59.5 to 36.0 on the Noometry Index. gpt-oss-120b costs 80× less per token, which makes it the better buy when GPT-5.4's lead doesn't matter for your workload.
Last verified . 36 shared benchmarks.
Summary
- They share 36 benchmarks with published results for both. GPT-5.4 scores higher in 9 categories and gpt-oss-120b in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 leads 61.7 to 19.5.
- The biggest single-benchmark swing is Terminal-Bench: 81.8% for GPT-5.4 and 18.7% for gpt-oss-120b.
- gpt-oss-120b is cheaper at $0.037 / $0.17 per million input/output tokens, against $2.50 / $15 for GPT-5.4.
- GPT-5.4 accepts more context: 1.05M tokens versus 131K.
- gpt-oss-120b has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 | gpt-oss-120b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 59.5 | 36.0 |
| Released | 2026-03-05 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 41K |
| Input $ / M tokens | $2.50 | $0.037 |
| Output $ / M tokens | $15 | $0.17 |
| Results tracked | 68 | 48 |
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-oss-120b: 33.2 (#251)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| SciCode | 56.6% | 36% |
| WeirdML | 77.7% | 48.2% |
| LMArena Coding | 1497 | 1380 |
| ALE-Bench | 1,607 | 575.62 |
| AlgoTune | 1.85 | 1.41 |
| SWE-bench Verified | 76.9% | — |
| DeepSWE | 51.8% | — |
| SWE-bench Verified (bash only) | — | 26% |
| Aider Polyglot | — | 41.8% |
| LMArena WebDev | 1465 | — |
| GSO | 31.4% | — |
| MirrorCode | 15.6% | — |
Agentic & Tool Use GPT-5.4 leads
GPT-5.4: 46.5 (#12), gpt-oss-120b: 12.1 (#145)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| Terminal-Bench | 81.8% | 18.7% |
| APEX-Agents | 52.4% | 4.4% |
| METR Time Horizons | 74.3% | 56.6% |
| Vending-Bench 2 | 6,144 | -21.53 |
| τ²-bench Banking | 39.4% | — |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | 19% | — |
| GBAEval | 45.1% | — |
| LMArena Search | 1197 | — |
Reasoning GPT-5.4 leads
GPT-5.4: 61.7 (#17), gpt-oss-120b: 19.5 (#236)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| Kagi LLM Benchmark | 63.8% | 58.6% |
| CritPt | 23.4% | 1.1% |
| Chess Puzzles | 44% | 20% |
| LMArena Hard Prompts | 1485 | 1364 |
| Mystery Game Puzzles | 37% | 2% |
| DTBench | 94.4% | 76.3% |
| LMCA | 52% | 22.1% |
| Epoch Capabilities Index | 156.92 | 139.94 |
| ARC-AGI-2 | 74% | — |
| SimpleBench | — | 22.1% |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 93.7% | — |
| EnigmaEval | 16% | — |
| Thematic Generalization | 80% | — |
| EBR-Bench | 25.4% | — |
| Surface Evolver Bench | — | 25% |
| ForecastBench | 59.5 | — |
Math GPT-5.4 leads
GPT-5.4: 73.3 (#14), gpt-oss-120b: 50.6 (#43)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 97.8% | 88.9% |
| LMArena Math | 1488 | 1388 |
| FrontierMath (Tiers 1-3) | 78.6% | — |
| FrontierMath Tier 4 | 49% | — |
| MathArena Final-Answer Competitions | 83.1% | — |
| ProofBench | 56% | — |
| Omni-MATH | — | 68.8% |
| FrontierMath (Feb 2025 set) | 47.6% | — |
| FrontierMath Tier 4 (v1) | 27.1% | — |
Knowledge GPT-5.4 leads
GPT-5.4: 65.3 (#13), gpt-oss-120b: 42.3 (#89)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| GPQA Diamond | 93.3% | 75.8% |
| Vectara Hallucination Rate | 7% | 14.2% |
| LMArena Expert | 1507 | 1356 |
| Humanity's Last Exam | 36.2% | — |
| SimpleQA Verified | 45.1% | — |
| MMLU-Pro | — | 79.5% |
| Confabulations | — | 15.7% |
| GPQA (HELM) | — | 68.4% |
Multimodal Not comparable
GPT-5.4: 43.8 (#19), gpt-oss-120b: —
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| LMArena Vision | 1302 | — |
| Blueprint-Bench 2 | 27.1% | — |
| Furniture Assembly | 37.5% | — |
| LMArena Document | 1471 | — |
Multilingual GPT-5.4 leads
GPT-5.4: 56.4 (#22), gpt-oss-120b: 48.2 (#142)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| LMArena Non-English | 1465 | 1350 |
| LMArena Chinese | 1520 | 1384 |
| LMArena French | 1493 | 1370 |
| LMArena German | 1472 | 1355 |
| LMArena Japanese | 1484 | 1332 |
| LMArena Korean | 1449 | 1283 |
| LMArena Russian | 1480 | 1343 |
| LMArena Spanish | 1453 | 1388 |
Instruction Following GPT-5.4 leads
GPT-5.4: 77.3 (#25), gpt-oss-120b: 69.5 (#168)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| LMArena Instruction Following | 1469 | 1318 |
| IFEval | — | 83.6% |
Long Context GPT-5.4 leads
GPT-5.4: 50.3 (#8), gpt-oss-120b: 31.4 (#273)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| LMArena Longer Query | 1474 | 1319 |
| Fiction.LiveBench | — | 44.4% |
| CL-bench | 27.9% | — |
| CL-bench Life | 21.7% | — |
Writing & Preference GPT-5.4 leads
GPT-5.4: 72.1 (#18), gpt-oss-120b: 46.7 (#213)
| Benchmark | GPT-5.4 | gpt-oss-120b |
|---|---|---|
| LMArena Text | 1469 | 1366 |
| LMArena Creative Writing | 1439 | 1274 |
| EQ-Bench Creative Writing | 1840 | 961 |
| LMArena Multi-Turn | 1482 | 1340 |
| Short-Story Creative Writing | — | 77.1% |
| WildBench | — | 84.5% |
| EQ-Bench 4 | 1272 | — |
Frequently asked questions
Is GPT-5.4 better than gpt-oss-120b?
GPT-5.4 is the stronger model overall, scoring 59.5 to 36.0 on the Noometry Index. gpt-oss-120b costs 80× 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-oss-120b?
gpt-oss-120b is cheaper. It lists at $0.037 per million input tokens and $0.17 per million output tokens; GPT-5.4 lists at $2.50 and $15.
Is GPT-5.4 or gpt-oss-120b better for coding?
GPT-5.4 scores higher on coding benchmarks: 52.8 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 does, with 1.05M tokens against 131K.
How many benchmarks do GPT-5.4 and gpt-oss-120b share?
36 benchmarks have published results for both models. GPT-5.4 has 68 scored results on Noometry and gpt-oss-120b has 48.