Model comparison
GPT-4.1 vs gpt-oss-20b
GPT-4.1 is the stronger model overall, scoring 35.7 to 32.1 on the Noometry Index. gpt-oss-20b costs 97× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Last verified . 31 shared benchmarks.
Summary
- They share 31 benchmarks with published results for both. GPT-4.1 scores higher in 6 categories and gpt-oss-20b in 3 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where GPT-4.1 leads 34.4 to 9.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 38.3% for GPT-4.1 and 65.3% for gpt-oss-20b.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $2 / $8 for GPT-4.1.
- GPT-4.1 accepts more context: 1.05M tokens versus 131K.
- gpt-oss-20b has downloadable open weights; the other is API-only.
Side by side
| GPT-4.1 | gpt-oss-20b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 35.7 | 32.1 |
| Released | 2025-04-14 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 33K | 16K |
| Input $ / M tokens | $2 | $0.018 |
| Output $ / M tokens | $8 | $0.09 |
| Results tracked | 52 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-20b leads
GPT-4.1: 34.2 (#228), gpt-oss-20b: 37.1 (#185)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| WeirdML | 39% | 40.9% |
| LMArena Coding | 1391 | 1307 |
| ALE-Bench | 558.1 | 566.05 |
| SWE-bench Verified | 48.5% | — |
| SWE-bench Verified (bash only) | 39.6% | — |
| Aider Polyglot | 52.4% | — |
| SciCode | — | 34.4% |
| CadEval | 42% | — |
Agentic & Tool Use GPT-4.1 leads
GPT-4.1: 34.4 (#40), gpt-oss-20b: 9.2 (#146)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| Terminal-Bench | — | 3.4% |
| Berkeley Function Calling Leaderboard | 54% | — |
Reasoning gpt-oss-20b leads
GPT-4.1: 11.7 (#334), gpt-oss-20b: 18.8 (#251)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 53.2% |
| Chess Puzzles | 6% | 4% |
| LMArena Hard Prompts | 1384 | 1274 |
| DTBench | 68.3% | 68% |
| LMCA | 25.6% | 14.5% |
| Epoch Capabilities Index | 136.79 | 137.81 |
| ARC-AGI-2 | 0.4% | — |
| SimpleBench | 27% | — |
| ARC-AGI-1 | 5.5% | — |
| CritPt | — | 1.4% |
| EnigmaEval | 2.2% | — |
| ForecastBench | 61.5 | — |
Math gpt-oss-20b leads
GPT-4.1: 21.4 (#269), gpt-oss-20b: 37.6 (#93)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 38.3% | 65.3% |
| Omni-MATH | 47.1% | 56.5% |
| LMArena Math | 1369 | 1317 |
| FrontierMath (Tiers 1-3) | 6% | — |
| MATH Level 5 | 83% | — |
| FrontierMath (Feb 2025 set) | 5.5% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge GPT-4.1 leads
GPT-4.1: 37.0 (#152), gpt-oss-20b: 34.6 (#186)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| GPQA Diamond | 66.9% | 60.8% |
| MMLU-Pro | 81.1% | 74% |
| GPQA (HELM) | 65.9% | 59.4% |
| LMArena Expert | 1364 | 1258 |
| Humanity's Last Exam | 5.4% | — |
| SimpleQA Verified | 31.1% | — |
| Vectara Hallucination Rate | 5.6% | — |
Multimodal Not comparable
GPT-4.1: 37.5 (#66), gpt-oss-20b: —
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| LMArena Vision | 1210 | — |
| GeoBench | 72% | — |
Multilingual GPT-4.1 leads
GPT-4.1: 49.6 (#128), gpt-oss-20b: 42.4 (#192)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| LMArena Non-English | 1370 | 1269 |
| LMArena Chinese | 1382 | 1311 |
| LMArena German | 1380 | 1249 |
| LMArena Japanese | 1320 | 1245 |
| LMArena Korean | 1339 | 1239 |
| LMArena Russian | 1377 | 1279 |
| LMArena Spanish | 1375 | 1267 |
| LMArena French | 1383 | — |
Instruction Following GPT-4.1 leads
GPT-4.1: 71.5 (#148), gpt-oss-20b: 62.0 (#235)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| IFEval | 83.8% | 73.2% |
| LMArena Instruction Following | 1367 | 1236 |
Long Context GPT-4.1 leads
GPT-4.1: 40.0 (#158), gpt-oss-20b: 37.9 (#203)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| LMArena Longer Query | 1385 | 1251 |
| Fiction.LiveBench | 63.9% | — |
Writing & Preference GPT-4.1 leads
GPT-4.1: 57.8 (#120), gpt-oss-20b: 35.8 (#260)
| Benchmark | GPT-4.1 | gpt-oss-20b |
|---|---|---|
| LMArena Text | 1383 | 1288 |
| LMArena Creative Writing | 1363 | 1201 |
| EQ-Bench Creative Writing | 1420 | 666 |
| WildBench | 85.4% | 73.7% |
| LMArena Multi-Turn | 1398 | 1268 |
Frequently asked questions
Is GPT-4.1 better than gpt-oss-20b?
GPT-4.1 is the stronger model overall, scoring 35.7 to 32.1 on the Noometry Index. gpt-oss-20b costs 97× less per token, which makes it the better buy when GPT-4.1's lead doesn't matter for your workload.
Which is cheaper, GPT-4.1 or gpt-oss-20b?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; GPT-4.1 lists at $2 and $8.
Is GPT-4.1 or gpt-oss-20b better for coding?
gpt-oss-20b scores higher on coding benchmarks: 37.1 versus 34.2 in the Noometry coding category.
Which has the bigger context window?
GPT-4.1 does, with 1.05M tokens against 131K.
How many benchmarks do GPT-4.1 and gpt-oss-20b share?
31 benchmarks have published results for both models. GPT-4.1 has 52 scored results on Noometry and gpt-oss-20b has 34.