Flagship OpenAI models:

3.5

4

4o

o1

o3

4.5

4.1

o4

5

5.1

5.2

5.3

words (e.g. thinking, codex, mini, max, pro, etc.) don’t count unless the abbreviation is what becomes synonymous with the model (e.g. no one calls 4o, 4 Omni)

the motivation for this market is seeing OpenAI 4 models based on GPT-5 and I’m wondering if they’ll release 5.4 all the way up to 5.9 where their increments are now by 0.1 every month or two

instead of sticking with a model for a year and adding checkpoints every few months

for example can we expect

Feb: GPT-5.3

March: GPT-5.4 / GPT-5.5

June: GPT-5.6

July: GPT-5.7

September: GPT-5.8

etc.

until GPT-6 (before GTA 6?)

will they add a random letter like k or p similar to Apple using s or e,

I’m wondering if I should allow 2 letters like SE or oss (that wasn’t a flagship, but just as an example), but I’m pretty SE means special edition (it could be 5s economy, which is why 16e) and oss is open source software, also if they use X like iPhone but pronounce it 10, the point of the market is to see how long until they break the numbering scheme or launch a new series similar to my iPhone market: https://manifold.markets/ChinmayTheMathGuy/by-which-year-will-apple-change-its

Criteria should be clear, but as long as OpenAI says the model is flagship it counts even if in reality there may be tradeoffs/cost cutting (Opus 4.6 vs 4.5, GPT-5 vs o3, 4o vs GPT-4, and even 4.1 vs 4.5 as long as there’s an argument and OpenAI advertises it as better it counts)

Resolution criteria

This market resolves based on which flagship OpenAI models are released before the introduction of a letter suffix (like 'o' in 4o, o1, or o3). GPT-5 was launched on August 7, 2025, and GPT-5.1 launched in November 2025 but was superseded almost immediately by GPT-5.2 in December.

A model counts as a flagship release if OpenAI officially designates it as such. Suffixes like "mini," "nano," "thinking," "codex," "pro," or other descriptors do not disqualify a model—only letter additions (e.g., 5a, 5k, 5p) would mark the transition. The market resolves YES for each model number that OpenAI releases as a flagship before adding a letter designation to the naming scheme.

Resolution sources: OpenAI's official models page and official OpenAI announcements via their website and API documentation.

Background

GPT-5 supersedes earlier AI models GPT-4, GPT-4o, GPT-4.1, GPT-4.5, and variants such as o3 and o4-mini. OpenAI has demonstrated a pattern of releasing incremental versions: after GPT-4, they released 4o (omni), 4.1, and 4.5 before moving to GPT-5. The current trajectory shows sub-version releases (5.1, 5.2) occurring within months of the base model launch, suggesting potential for continued 0.1 increments.

Considerations

OpenAI's naming conventions have been inconsistent and frequently criticized. There are "crimes against consistent naming conventions, capitalization, and hyphenation" in their model lineup. The company may introduce letter suffixes unpredictably or skip certain version numbers entirely. Additionally, the distinction between what OpenAI markets as "flagship" versus specialized variants (Codex, thinking models, etc.) can be ambiguous, though the user's criteria clarifies that OpenAI's own designation determines flagship status.

This description was generated by AI.