gpt-oss-120b
OpenAI
Key facts · verified Oct 29, 2026
gpt-oss-120b is an OpenAI model released on 5 Aug 2025. Its API price is $0.03 per million input tokens and $0.17 per million output tokens (OpenRouter). Its context window is 131k tokens. It ranks #219 on LMArena's text leaderboard (Elo 1352, read 2 Nov 2026). Source: huggingface.co.
OpenAI's open-weight Mixture-of-Experts model (117B total, 5.1B active), released under Apache 2.0. OpenAI does not run a first-party paid API for it; pricing below is from OpenRouter, one of several third-party hosts.
Pricing & specs checkedSource: vendor documentation · checked Oct 29, 2026LMArena: 1352 Elo (rank 219) · source · retrieved Nov 2, 2026
Key Features
Open Weights Configurable Reasoning Tool Use Structured Output Self-Hostable
Full Specifications
$0.03 / $0.17 per million tokens | |
|---|---|
| Identity | |
| Developer | OpenAI |
| Released | Aug 2025 |
| Status | GA |
| Licence | Apache 2.0 |
| Self-hostable | Yes |
| Cost | |
| Blended $/1M tokensinput × 0.75 + output × 0.25 | $0.065 / 1M tokens |
| Input price | $0.030 / 1M tokens |
| Output price | $0.170 / 1M tokens |
| Cached input | Varies by host |
| Batch discount | Varies by host |
| Free tier | Free tiers vary by third-party host |
| Capacity | |
| Context window | 131k tokens |
| Max output | 131k tokens |
| Long-context surcharge | Not applicable |
| Capability | |
| Vision in | No |
| Audio in | No |
| Function calling | Yes |
| Structured output | Yes |
| Extended reasoning | Yes (configurable reasoning effort) |
| Web search | No |
| Code execution | No |
| Access | |
| API | Via third-party hosts |
| Chat app | No (open weights) |
| Cloud marketplaces | Not applicable (open weights) |
| Fine-tuning | Yes (open weights) |
| Measured quality | |
| LMArena Elo | 1352 (checked Nov 2026) |
| LMArena rank | Rank 219 |
| Elo per dollarLMArena Elo ÷ blended $/1M tokens | 20800 |