← Founder Notes
Archive

Openai released gpt-oss-120b and gpt-oss-20b under apache 2.0 — its first open weights since gpt-2,…

Yethikrishna ROriginal on Threads

openai released gpt-oss-120b and gpt-oss-20b under apache 2.0 — its first open weights since gpt-2, closing a five-year closed era. the same company selling agentic subscriptions just gave away its flagship-class models.

the moat was never the weights.

Context

OpenAI's gpt-oss post, dated 5 August 2025, introduces gpt-oss-120b and gpt-oss-20b as open-weight models under Apache 2.0. It says the 120b is near parity with o4-mini on core reasoning benchmarks and runs on a single 80 GB GPU, and the 20b is similar to o3-mini and needs 16 GB of memory. OpenAI's open models page also lists gpt-oss-safeguard.

How it compares

The first-party release is 5 August 2025, a year before the note, which gives no date. OpenAI's post calls these its first open-weight language models since GPT-2, which supports that part; a five-year closed era and the year 2019 are not in the OpenAI text read, and the elapsed time is the author's claim. OpenAI compares the models with o4-mini and o3-mini, not its flagships, so flagship-class is not supported.

Related work

Watch next

  • Later OpenAI open-weight releases.

Sources

  1. Introducing gpt-oss (OpenAI, 5 Aug 2025)openai.com

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 18 September 2026 at 07:52 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/openai-released-gpt-oss-120b-and-gpt-oss-DdaWicEjFpz" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="Openai released gpt-oss-120b and gpt-oss-20b under apache 2.0 — its first open weights since gpt-2,…"></iframe>

More notes