OpenAI: gpt-oss-120b

GPT-OSS-120B is a text-in, text-out model from OpenAI with a 131,072-token context window. It supports tool use and reasoning, which makes it usable for multi-step agentic workflows, though structured output support is unconfirmed from available data. There is no documented completion token cap in the current spec, so users should verify output limits with the provider before committing to long-generation tasks. At $0.039 per million input tokens and $0.18 per million output tokens, the pricing sits at the budget end of the reasoning-capable tier. The blended benchmark score of 42.3 across only 5 benchmarks is a thin evidence base, so treat performance claims cautiously. Coding tasks show the strongest result at 50.2, while agentic performance at 21.7 is notably weak. Teams doing cost-sensitive coding work may find it worth a trial, but those prioritizing agentic reliability or needing broader benchmark validation should compare alternatives before deciding.