Model comparison
gpt-oss-20b vs o3-mini
o3-mini is the stronger model overall, scoring 36.7 to 32.5 on the Noometry Index. gpt-oss-20b costs 53× less per token, which makes it the better buy when o3-mini's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
Summary
- They share 25 benchmarks with published results for both. gpt-oss-20b scores higher in 3 categories and o3-mini in 6 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where o3-mini leads 29.6 to 9.3.
- The biggest single-benchmark swing is GPQA Diamond: 60.8% for gpt-oss-20b and 77% for o3-mini.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $1.10 / $4.40 for o3-mini.
- o3-mini accepts more context: 200K tokens versus 131K.
- gpt-oss-20b has downloadable open weights; the other is API-only.
Side by side
| gpt-oss-20b | o3-mini | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 32.5 | 36.7 |
| Released | 2025-08-05 | 2024-12-20 |
| Weights | Open | Proprietary |
| Context window | 131K | 200K |
| Max output | 16K | 100K |
| Input $ / M tokens | $0.018 | $1.10 |
| Output $ / M tokens | $0.09 | $4.40 |
| Results tracked | 34 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding o3-mini leads
gpt-oss-20b: 37.6 (#192), o3-mini: 40.8 (#132)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| SciCode | 34.4% | 39.8% |
| WeirdML | 40.9% | 43.7% |
| LMArena Coding | 1306 | 1378 |
| Aider Polyglot | — | 60.4% |
| GSO | — | 1.3% |
| LiveBench Coding | — | 82.7% |
| CadEval | — | 54% |
| ALE-Bench | 566.05 | — |
Agentic & Tool Use o3-mini leads
gpt-oss-20b: 9.3 (#154), o3-mini: 29.6 (#84)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| Terminal-Bench | 3.4% | — |
| Cybench | — | 22.5% |
Reasoning gpt-oss-20b leads
gpt-oss-20b: 19.3 (#261), o3-mini: 16.3 (#305)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| CritPt | 1.4% | 0.3% |
| Chess Puzzles | 4% | 17% |
| LMArena Hard Prompts | 1274 | 1366 |
| DTBench | 68% | 68.8% |
| LMCA | 14.5% | 19% |
| Epoch Capabilities Index | 137.82 | 140.34 |
| ARC-AGI-2 | — | 3% |
| SimpleBench | — | 22.8% |
| Kagi LLM Benchmark | 53.2% | — |
| ARC-AGI-1 | — | 34.5% |
| LiveBench Reasoning | — | 89.6% |
| Mystery Game Puzzles | — | 7% |
| LiveBench Data Analysis | — | 70.6% |
| ForecastBench | — | 59.6 |
| LiveBench | — | 75.9% |
Math gpt-oss-20b leads
gpt-oss-20b: 39.4 (#103), o3-mini: 28.1 (#244)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | 76.9% |
| LMArena Math | 1317 | 1396 |
| FrontierMath (Tiers 1-3) | — | 18.6% |
| FrontierMath Tier 4 | — | 0% |
| Omni-MATH | 56.5% | — |
| LiveBench Math | — | 77.3% |
| MATH Level 5 | — | 96.5% |
| FrontierMath (Feb 2025 set) | — | 12.4% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge o3-mini leads
gpt-oss-20b: 34.6 (#195), o3-mini: 38.3 (#146)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| GPQA Diamond | 60.8% | 77% |
| LMArena Expert | 1258 | 1364 |
| SimpleQA Verified | — | 15.3% |
| MMLU-Pro | 74% | — |
| Confabulations | — | 17.9% |
| GPQA (HELM) | 59.4% | — |
Multilingual o3-mini leads
gpt-oss-20b: 42.2 (#197), o3-mini: 45.7 (#164)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| LMArena Non-English | 1268 | 1319 |
| LMArena Chinese | 1314 | 1379 |
| LMArena German | 1255 | 1303 |
| LMArena Japanese | 1244 | 1286 |
| LMArena Korean | 1236 | 1314 |
| LMArena Russian | 1278 | 1304 |
| LMArena Spanish | 1267 | 1321 |
| LMArena French | — | 1334 |
Instruction Following o3-mini leads
gpt-oss-20b: 61.8 (#240), o3-mini: 75.1 (#72)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| LMArena Instruction Following | 1236 | 1337 |
| LiveBench Instruction Following | — | 84.4% |
| IFEval | 73.2% | — |
Long Context gpt-oss-20b leads
gpt-oss-20b: 37.9 (#209), o3-mini: 33.8 (#256)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| LMArena Longer Query | 1250 | 1343 |
| Fiction.LiveBench | — | 50% |
Writing & Preference o3-mini leads
gpt-oss-20b: 35.5 (#265), o3-mini: 50.3 (#182)
| Benchmark | gpt-oss-20b | o3-mini |
|---|---|---|
| LMArena Text | 1287 | 1337 |
| LMArena Creative Writing | 1201 | 1286 |
| LMArena Multi-Turn | 1268 | 1320 |
| Short-Story Creative Writing | — | 61.7% |
| EQ-Bench Creative Writing | 666 | — |
| WildBench | 73.7% | — |
| LiveBench Language | — | 50.7% |
Frequently asked questions
Is gpt-oss-20b better than o3-mini?
o3-mini is the stronger model overall, scoring 36.7 to 32.5 on the Noometry Index. gpt-oss-20b costs 53× less per token, which makes it the better buy when o3-mini's lead doesn't matter for your workload.
Which is cheaper, gpt-oss-20b or o3-mini?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; o3-mini lists at $1.10 and $4.40.
Is gpt-oss-20b or o3-mini better for coding?
o3-mini scores higher on coding benchmarks: 40.8 versus 37.6 in the Noometry coding category.
Which has the bigger context window?
o3-mini does, with 200K tokens against 131K.
How many benchmarks do gpt-oss-20b and o3-mini share?
25 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and o3-mini has 51.