GPT-6 Astra

Mon, 01 Jan 0001 00:00:00 +0000

GPT-6 Astra is OpenAI’s current flagship, released 3 September 2026 under the identifier gpt-6-astra. It takes text and images and returns text, with a 1,050,000-token context window and a reasoning effort setting running from low to max.

Two things about it are unusual enough to be the reason for this page. It is a single model rather than a tiered family, breaking the naming scheme OpenAI had introduced one generation earlier. And its system card declares it the first OpenAI model to reach the Critical cybersecurity level under the company’s own risk framework — a threshold that, by OpenAI’s account, changed how the model is secured internally as well as how it is released.

gpt-oss

Mon, 01 Jan 0001 00:00:00 +0000

gpt-oss is a pair of open-weight models OpenAI released in August 2025 under Apache 2.0 — gpt-oss-120b and gpt-oss-20b. They are the company’s first open-weight language models since GPT-2 in 2019, and they come with a model card that describes the architecture completely: layer counts, expert counts, routing width, attention configuration, quantisation format.

That makes them unusual in a specific way. Every OpenAI flagship since GPT-4 has been covered by an explicit refusal to disclose architecture. These two are not. For anyone who wants to see what a current-generation model from a frontier lab actually looks like inside, gpt-oss is the only OpenAI answer — and where GPT-2 shows the 2019 design running on a laptop, this shows what changed in six years.

OpenAI on Bitsy Wiki

GPT-6 Astra

gpt-oss