Model comparison
gpt-oss-20b vs Phi-4 Mini
gpt-oss-20b is the stronger model overall, scoring 32.5 to 30.9 on the Noometry Index.
Last verified . 2 shared benchmarks.
MicrosoftSummary
- They share 2 benchmarks with published results for both. gpt-oss-20b scores higher in 2 categories and Phi-4 Mini in 1 category; 3 gaps are clear of the uncertainty.
- The widest gap is in coding, where gpt-oss-20b leads 37.6 to 28.1.
- The biggest single-benchmark swing is SciCode: 34.4% for gpt-oss-20b and 10.8% for Phi-4 Mini.
- gpt-oss-20b is cheaper at $0.018 / $0.09 per million input/output tokens, against $0.075 / $0.30 for Phi-4 Mini.
- gpt-oss-20b accepts more context: 131K tokens versus 128K.
Side by side
| gpt-oss-20b | Phi-4 Mini | |
|---|---|---|
| Provider | OpenAI | Microsoft |
| Noometry Index | 32.5 | 30.9 |
| Released | 2025-08-05 | 2024-12-11 |
| Weights | Open | Open |
| Context window | 131K | 128K |
| Max output | 16K | 4K |
| Input $ / M tokens | $0.018 | $0.075 |
| Output $ / M tokens | $0.09 | $0.30 |
| Results tracked | 34 | 3 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding gpt-oss-20b leads
gpt-oss-20b: 37.6 (#192), Phi-4 Mini: 28.1 (#317)
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| SciCode | 34.4% | 10.8% |
| WeirdML | 40.9% | — |
| LMArena Coding | 1306 | — |
| ALE-Bench | 566.05 | — |
Agentic & Tool Use Not comparable
gpt-oss-20b: 9.3 (#154), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| Terminal-Bench | 3.4% | — |
Reasoning Phi-4 Mini leads
gpt-oss-20b: 19.3 (#261), Phi-4 Mini: 22.4 (#195)
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| CritPt | 1.4% | 0% |
| Kagi LLM Benchmark | 53.2% | — |
| Chess Puzzles | 4% | — |
| LMArena Hard Prompts | 1274 | — |
| DTBench | 68% | — |
| LMCA | 14.5% | — |
| Epoch Capabilities Index | 137.82 | — |
Math Not comparable
gpt-oss-20b: 39.4 (#103), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 65.3% | — |
| Omni-MATH | 56.5% | — |
| LMArena Math | 1317 | — |
Knowledge gpt-oss-20b leads
gpt-oss-20b: 34.6 (#195), Phi-4 Mini: 25.3 (#262)
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| GPQA Diamond | 60.8% | — |
| MMLU-Pro | 74% | — |
| Vectara Hallucination Rate | — | 23.5% |
| GPQA (HELM) | 59.4% | — |
| LMArena Expert | 1258 | — |
Multilingual Not comparable
gpt-oss-20b: 42.2 (#197), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| LMArena Non-English | 1268 | — |
| LMArena Chinese | 1314 | — |
| LMArena German | 1255 | — |
| LMArena Japanese | 1244 | — |
| LMArena Korean | 1236 | — |
| LMArena Russian | 1278 | — |
| LMArena Spanish | 1267 | — |
Instruction Following Not comparable
gpt-oss-20b: 61.8 (#240), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| IFEval | 73.2% | — |
| LMArena Instruction Following | 1236 | — |
Long Context Not comparable
gpt-oss-20b: 37.9 (#209), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| LMArena Longer Query | 1250 | — |
Writing & Preference Not comparable
gpt-oss-20b: 35.5 (#265), Phi-4 Mini: —
| Benchmark | gpt-oss-20b | Phi-4 Mini |
|---|---|---|
| LMArena Text | 1287 | — |
| LMArena Creative Writing | 1201 | — |
| EQ-Bench Creative Writing | 666 | — |
| WildBench | 73.7% | — |
| LMArena Multi-Turn | 1268 | — |
Frequently asked questions
Is gpt-oss-20b better than Phi-4 Mini?
gpt-oss-20b is the stronger model overall, scoring 32.5 to 30.9 on the Noometry Index.
Which is cheaper, gpt-oss-20b or Phi-4 Mini?
gpt-oss-20b is cheaper. It lists at $0.018 per million input tokens and $0.09 per million output tokens; Phi-4 Mini lists at $0.075 and $0.30.
Is gpt-oss-20b or Phi-4 Mini better for coding?
gpt-oss-20b scores higher on coding benchmarks: 37.6 versus 28.1 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-oss-20b and Phi-4 Mini share?
2 benchmarks have published results for both models. gpt-oss-20b has 34 scored results on Noometry and Phi-4 Mini has 3.