GPT-OSS-20B: gPQA Diamond now lists OpenAI’s published launch result of 71.5%; results from different evaluation settings are not interchangeable

The OnAirToday AI Progress Tracker recorded a verified change to OpenAI's GPT-OSS-20B (version gpt-oss-20b): GPQA Diamond now lists OpenAI’s published launch result of 71.5%; results from different evaluation settings are not interchangeable.

GPT-OSS-20B is OpenAI’s Apache 2.0–licensed, open-weight text reasoning model, supporting adjustable reasoning effort and tool use. Its released quantization targets deployment within 16 GB of memory, while actual serving requirements and hosted inference prices depend on the implementation and provider.

What changed in this review:

• Added OpenAI’s published SWE-bench Verified launch result of 60.7%.
• Removed the OTIS Mock AIME and Terminal-Bench entries; their exact results are not yet confirmed by public reporting.
• Removed fixed token prices because hosted pricing is provider-dependent, not an intrinsic price of the open-weight release.

Teams building on GPT-OSS-20B should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.

The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/gpt-oss-20b.

References:

• https://openai.com/index/introducing-gpt-oss/
• https://huggingface.co/openai/gpt-oss-20b
• https://github.com/openai/gpt-oss