Model comparison
GPT-4 Turbo vs gpt-oss-20b
gpt-oss-20b is the stronger model overall, scoring 32.5 to 30.5 on the Noometry Index.
Last verified . 23 shared benchmarks.
Summary
- They share 23 benchmarks with published results for both. GPT-4 Turbo scores higher in 3 categories and gpt-oss-20b in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where gpt-oss-20b leads 39.4 to 9.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 6.7% for GPT-4 Turbo and 65.3% for gpt-oss-20b.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $10 / $30 for GPT-4 Turbo.
- gpt-oss-20b accepts more context: 131K tokens versus 128K.
- gpt-oss-20b has downloadable open weights; the other is API-only.
Side by side
| GPT-4 Turbo | gpt-oss-20b | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 30.5 | 32.5 |
| Released | 2023-11-06 | 2025-08-05 |
| Weights | Proprietary | Open |
| Context window | 128K | 131K |
| Max output | 4K | 16K |
| Input $ / M tokens | $10 | $0.018 |
| Output $ / M tokens | $30 | $0.09 |
| Results tracked | 36 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-20b leads
GPT-4 Turbo: 33.8 (#249), gpt-oss-20b: 37.6 (#192)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| WeirdML | 18% | 40.9% |
| LMArena Coding | 1268 | 1306 |
| SciCode | — | 34.4% |
| BigCodeBench Instruct | 48.2% | — |
| BigCodeBench Complete | 58.2% | — |
| ALE-Bench | — | 566.05 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73.3% | — |
Agentic & Tool Use Not comparable
GPT-4 Turbo: —, gpt-oss-20b: 9.3 (#154)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| Terminal-Bench | — | 3.4% |
| METR Time Horizons | 36.7% | — |
Reasoning gpt-oss-20b leads
GPT-4 Turbo: 15.3 (#317), gpt-oss-20b: 19.3 (#261)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| Chess Puzzles | 6% | 4% |
| LMArena Hard Prompts | 1251 | 1274 |
| DTBench | 61.6% | 68% |
| LMCA | 9.8% | 14.5% |
| Epoch Capabilities Index | 127.25 | 137.82 |
| SimpleBench | 25.1% | — |
| Kagi LLM Benchmark | — | 53.2% |
| CritPt | — | 1.4% |
| ForecastBench | 59.4 | — |
Math gpt-oss-20b leads
GPT-4 Turbo: 9.0 (#322), gpt-oss-20b: 39.4 (#103)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 6.7% | 65.3% |
| LMArena Math | 1272 | 1317 |
| FrontierMath (Tiers 1-3) | 0.7% | — |
| Omni-MATH | — | 56.5% |
| MATH Level 5 | 46.7% | — |
Knowledge gpt-oss-20b leads
GPT-4 Turbo: 24.3 (#268), gpt-oss-20b: 34.6 (#195)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| GPQA Diamond | 46.6% | 60.8% |
| LMArena Expert | 1223 | 1258 |
| MMLU-Pro | — | 74% |
| Confabulations | 28.4% | — |
| GPQA (HELM) | — | 59.4% |
| MMLU | 81.3% | — |
Multimodal Not comparable
GPT-4 Turbo: 30.6 (#110), gpt-oss-20b: —
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| LMArena Vision | 1090 | — |
Multilingual gpt-oss-20b leads
GPT-4 Turbo: 40.5 (#216), gpt-oss-20b: 42.2 (#197)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| LMArena Non-English | 1245 | 1268 |
| LMArena Chinese | 1242 | 1314 |
| LMArena German | 1259 | 1255 |
| LMArena Japanese | 1194 | 1244 |
| LMArena Korean | 1187 | 1236 |
| LMArena Russian | 1259 | 1278 |
| LMArena Spanish | 1260 | 1267 |
| LMArena French | 1276 | — |
Instruction Following GPT-4 Turbo leads
GPT-4 Turbo: 65.8 (#216), gpt-oss-20b: 61.8 (#240)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| LMArena Instruction Following | 1249 | 1236 |
| IFEval | — | 73.2% |
Long Context Too close to call
GPT-4 Turbo: 38.0 (#206), gpt-oss-20b: 37.9 (#209)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| LMArena Longer Query | 1254 | 1250 |
Writing & Preference GPT-4 Turbo leads
GPT-4 Turbo: 47.7 (#206), gpt-oss-20b: 35.5 (#265)
| Benchmark | GPT-4 Turbo | gpt-oss-20b |
|---|---|---|
| LMArena Text | 1272 | 1287 |
| LMArena Creative Writing | 1269 | 1201 |
| LMArena Multi-Turn | 1267 | 1268 |
| EQ-Bench Creative Writing | — | 666 |
| WildBench | — | 73.7% |
Frequently asked questions
Is GPT-4 Turbo better than gpt-oss-20b?
gpt-oss-20b is the stronger model overall, scoring 32.5 to 30.5 on the Noometry Index.
Which is cheaper, GPT-4 Turbo 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 Turbo lists at $10 and $30.
Is GPT-4 Turbo or gpt-oss-20b better for coding?
gpt-oss-20b scores higher on coding benchmarks: 37.6 versus 33.8 in the Noometry coding category.
Which has the bigger context window?
gpt-oss-20b does, with 131K tokens against 128K.
How many benchmarks do GPT-4 Turbo and gpt-oss-20b share?
23 benchmarks have published results for both models. GPT-4 Turbo has 36 scored results on Noometry and gpt-oss-20b has 34.