# GPT-OSS 120B vs GPT-OSS 20B

> Side-by-side benchmark comparison for GPT-OSS 120B and GPT-OSS 20B across agentic, coding, multimodal, knowledge, reasoning, multilingual, and math tasks.

- Canonical page: https://benchlm.ai/compare/gpt-oss-120b-vs-gpt-oss-20b
- Last updated: October 17, 2026

- Shared sourced benchmarks: 7
- HTML indexing: indexable
- Ranking lane: BenchAlign v5.8

## Quick Verdict

Pick GPT-OSS 120B if you want the stronger benchmark profile. GPT-OSS 20B only makes more sense when its price, context window, or workload-specific category wins matter more than the overall score.

## Summary

- GPT-OSS 120B leads overall 37.32 to 33.32.
- The clearest category separation is in reasoning, where the averages are 61.6 for GPT-OSS 120B and 49.7 for GPT-OSS 20B.
- The biggest single benchmark swing is GPQA in Knowledge, with scores of 80.1% and 71.5%.

## Model Snapshot

| Property | GPT-OSS 120B | GPT-OSS 20B |
|----------|----------|----------|
| Creator | OpenAI | OpenAI |
| Type | Open Weight | Open Weight |
| Reasoning | Non-Reasoning | Non-Reasoning |
| Context | 128K | 128K |
| Overall Score | 37.32 | 33.32 |
| Benchmarks Covered | 8 | 7 |
| Pricing (input/output) | $0.00 / $0.00 | $0.00 / $0.00 |

## Category Breakdown

### Agentic

- Winner: Directional only
- GPT-OSS 120B public-lane score: 19.5 (Estimated · #110/128)
- GPT-OSS 20B public-lane score: 14.9 (Estimated · #115/128)

| Benchmark | GPT-OSS 120B | GPT-OSS 20B | Winner |
|-----------|-----------|-----------|--------|
| Gert Labs | 29.61% | Coming soon | Coming soon |

### Coding

- Winner: Directional only
- GPT-OSS 120B public-lane score: 23 (Estimated · #111/150)
- GPT-OSS 20B public-lane score: 18.2 (Estimated · #123/150)

| Benchmark | GPT-OSS 120B | GPT-OSS 20B | Winner |
|-----------|-----------|-----------|--------|
| SWE-bench Verified | 62.4% | 60.7% | GPT-OSS 120B |
| React Native Evals | 71.6% | 71% | GPT-OSS 120B |

### Reasoning

- Winner: Not comparable
- GPT-OSS 120B public-lane score: 61.6 (Unranked · 2 rankable rows)
- GPT-OSS 20B public-lane score: 49.7 (Unranked · 2 rankable rows)

### Knowledge

- Winner: GPT-OSS 120B
- GPT-OSS 120B public-lane score: 38.6 (Supported · #122/184)
- GPT-OSS 20B public-lane score: 33.9 (Supported · #140/184)

| Benchmark | GPT-OSS 120B | GPT-OSS 20B | Winner |
|-----------|-----------|-----------|--------|
| MMLU | 90% | 85.3% | GPT-OSS 120B |
| GPQA | 80.1% | 71.5% | GPT-OSS 120B |
| HLE | 14.9% | 10.9% | GPT-OSS 120B |

### Instruction Following

- Winner: Directional only
- GPT-OSS 120B public-lane score: 84.3 (#50/148)
- GPT-OSS 20B public-lane score: 79.2 (#60/148)

### Mathematics

- Winner: Not comparable
- GPT-OSS 120B public-lane score: Coming soon
- GPT-OSS 20B public-lane score: Coming soon

| Benchmark | GPT-OSS 120B | GPT-OSS 20B | Winner |
|-----------|-----------|-----------|--------|
| AIME 2024 | 95.8% | 92.1% | GPT-OSS 120B |
| AIME 2025 | 92.5% | 91.7% | GPT-OSS 120B |

## FAQ

### Which is better overall, GPT-OSS 120B or GPT-OSS 20B?

GPT-OSS 120B is ahead overall on BenchLM right now.

### Where is the biggest gap between GPT-OSS 120B and GPT-OSS 20B?

The widest category gap is in reasoning, where the averages are 61.6 for GPT-OSS 120B and 49.7 for GPT-OSS 20B.

### How many benchmarks does GPT-OSS 120B cover on BenchLM?

GPT-OSS 120B currently has 8 sourced benchmark scores on BenchLM.

### How many benchmarks does GPT-OSS 20B cover on BenchLM?

GPT-OSS 20B currently has 7 sourced benchmark scores on BenchLM.

## Related Comparisons

- [GPT-OSS 120B vs Claude Opus 5.5](/compare/claude-opus-5-5-vs-gpt-oss-120b)
- [GPT-OSS 20B vs Claude Opus 5.5](/compare/claude-opus-5-5-vs-gpt-oss-20b)
- [GPT-OSS 120B vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-gpt-oss-120b)
- [GPT-OSS 20B vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-gpt-oss-20b)

## Explore More

- [GPT-OSS 120B profile](/models/gpt-oss-120b)
- [GPT-OSS 20B profile](/models/gpt-oss-20b)
- [Compare Pricing](/llm-pricing)
- [Alternative Finder](/tools/alternative-finder)
- [LLM Selector](/tools/llm-selector)
- [Overall Rankings](/best/overall)