Calls for a sequel to OpenAI's open-weight LLM family "gpt-oss"—dubbed "gpt-oss2"—are spreading among developers. As of June 2026, however, no official announcement or release from OpenAI has been confirmed.
June 15, 2026 · OpenAI open-weight models
Developers keep asking for "gpt-oss2" — but no sequel is confirmed
After OpenAI's first open-weight release in August 2025 won praise for strong reasoning and easy local use, the community's push for a follow-up has carried into June 2026. As of now, gpt-oss2 remains request and speculation — with no official announcement.
2
Models released under Apache 2.0: gpt-oss-120b & gpt-oss-20b
128k
Context length supported by both models
6 yrs
Since OpenAI's last open weights (GPT-2, 2019)
Total parameters — the two models to scale
Both use a Mixture-of-Experts design; only a fraction is active per token.
117B
gpt-oss-120b
5.1B active / token
21B
gpt-oss-20b
3.6B active / token
What developers praise
Strong reasoning & agentic / tool-use performance
Visible Chain-of-Thought reasoning
Easy local run & fine-tune (Ollama, LM Studio)
"Best-in-class for local use"
Why a sequel is wanted
"Excellent, but leaves room for a next step"
Pressure from rivals like Qwen and DeepSeek
Speculation: a move once a new DeepSeek lands
Still no official gpt-oss2 confirmation
Bottom line
The original gpt-oss models remain popular for local deployment — but as of June 2026, "gpt-oss2" exists only as community requests and speculation, not as an OpenAI release.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…