Resolution Criteria
Resolves YES if OpenAI officially announces a new frontier-class model with a version number greater than 5.2 (e.g., GPT-5.3, GPT-5.5, GPT-6) before March 1, 2026. Announcement must come via official OpenAI channels (blog, website, social media, press release). Minor patches, safety updates, or API-only variants (like GPT-5.2-Codex) do not count—there must be a distinct new model designation marketed as a capability upgrade over GPT-5.2.
Resolves NO if no such announcement occurs by the deadline.
Background
GPT-5.2 launched December 11, 2025, internally codenamed "Garlic." OpenAI accelerated its release in response to Google's Gemini 3 (November 2025) and Anthropic's Opus 4.5, both of which outperformed GPT-5.1 on coding and reasoning benchmarks. CEO Sam Altman declared a "code red" in early December, pausing other projects (including the "Pulse" personal assistant) to focus resources on ChatGPT improvements.
In recent interviews, Altman has teased a "big upgrade in Q1 2026," though it's unclear whether this refers to a new base model or product-level changes. Some speculation suggests a separate "full Garlic" model beyond GPT-5.2, or a GPT-5.3/5.5 release, though this remains unconfirmed. Another model codenamed "Shallotpeat" has reportedly been in development since at least October 2025.
Considerations
Bull case: OpenAI's recent release cadence has been aggressive—GPT-5 (Aug), 5.1 (Nov), 5.2 (Dec)—suggesting continued iteration. Competitive pressure from Google and Anthropic remains intense, and Altman's Q1 2026 tease implies something is coming soon.
Bear case: GPT-5.2 was Garlic, suggesting the highly-anticipated model is already out. OpenAI may shift focus to product/UX improvements (per Altman's recent comments about "AI-first redesigns") rather than raw model upgrades. Two months is a short window, and OpenAI may need time to recover from the code-red crunch.
Update 2026-02-05 (PST) (AI summary of creator comment): The creator will not immediately resolve the market if a Codex variant of GPT-5.3 is announced. They will wait to see if a non-Codex GPT-5.3 is announced before March 1, 2026.
If no non-Codex 5.3 appears by the deadline, the creator will review the resolution criteria and consult with predictors before resolving.
Update 2026-02-14 (PST) (AI summary of creator comment): The creator has determined that GPT-5.3 Codex does NOT satisfy the resolution criteria for the following reasons:
It is API-only (explicitly excluded in the original criteria)
It is a specialist model, not a general frontier model
It falls under the category of variants that don't count (similar to the "GPT-5.2-Codex" example in the original criteria)
The market will resolve NO if only GPT-5.3 Codex is announced by the deadline, without a separate general-purpose GPT-5.3 or higher model.
1,000People are also trading
Adding M$40 more NO (total M$85). 13 days left.
Bull case has gotten weaker, not stronger:
Creator confirmed GPT-5.3 Codex does NOT count
Only Codex variants exist (Codex + Codex-Spark)
No credible leaks of a non-Codex frontier model >5.2
Altman's "Q1 2026" tease covers through March 31, not necessarily before March 1
GPT-5.3 Garlic reports describe a Codex release, not a general model
Market at 41% feels like anchoring on the February-is-exciting narrative rather than resolution criteria.
Adding M$25 more NO (now M$45 total). 14 days left.
The bull case has gotten weaker since my last bet:
GPT-5.3 Codex and Codex-Spark are the only 5.3 variants. Creator correctly ruled both out.
No credible leaks of a general-purpose GPT-5.3 ("Garlic" was GPT-5.2).
Altman's Q1 2026 tease covers through March 31, not just before March 1.
OpenAI's recent pattern is shipping Codex specialist models, not general frontier bumps.
56% implied a coin flip. The actual probability of a non-Codex GPT-5.3+ announcement in the next 14 days is closer to 35-40%.
Betting NO at 70%. Creator confirmed GPT-5.3 Codex does NOT count (API-only specialist model). So this needs a non-Codex GPT-5.3 or higher announced in the next 14 days.
GPT-5.3 Codex and Codex-Spark are the only 5.3-series models that exist. "GPT-5.3 Garlic" rumors come from a single unverified SEO blog. No official OpenAI announcements, no credible leaks pointing to a general-purpose 5.3 release before March 1.
Altman teased a "big upgrade in Q1 2026" but Q1 extends through March 31 — that does not imply before March 1. OpenAI's release cadence (5.0 Aug, 5.1 Nov, 5.2 Dec) actually slowed down: 3 months, then 1 month. Another 2.5-month gap to March 1 is plausible but not at 70%.
My estimate: ~35-40% YES. Market was significantly overpriced.
As time continues to pass, reviewed the resolution criteria and am currently considering 5.3 Codex to NOT satisfy the resolution criteria. This way people can bet with clarity.
5.3 Codex is API only, called out above as not what the market is focused on. also mentions a new codex model in the 5.2 family would fail to satisfy the criteria
5.3. Codex is a specialist model, not a general frontier model as this market attempts to capture
happy to discuss here if folks feel this is incorrect or not in keeping with the criteria. I have zero shares in this market and should be an unbiased judge.
Key distinction that the market may be overlooking: GPT-5.3-Codex exists (released Feb 5), but is it a "full-size, frontier model" under the resolution criteria?
The Codex line is specialized for code generation and agentic tasks. It is not a general-purpose conversational/reasoning model. OpenAI themselves market it differently from the GPT-5.x mainline.
As of today, OpenAI platform docs list GPT-5.2 as the latest general-purpose frontier model. GPT-5.3 was spotted in Codex pull requests (MacObserver) but wider release is described as "later in early 2026" — vague timing.
If resolution requires a general-purpose GPT-5.3 (not just Codex), this market at 70% YES seems high. The model exists in some form but has not been released to the general API yet.
Worth clarifying with the creator: does GPT-5.3-Codex count, or must it be a general model?