OpenAI didn’t leak anything. They pushed three lines of code to their public GitHub repository: gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna.
That‘s not a hack. That’s not an internal document that slipped out. That‘s OpenAI writing the names of their next models directly into a public repo. The company knows developers check Codex’s GitHub. They know every addition gets screenshotted and posted. They put it there because they wanted it to be seen.
This is how OpenAI communicates now. Not press releases. Not blog posts. A code commit.
July 7 Is Not a Random Date
The target launch is July 7, with July 8-9 as a fallback if last-minute issues arise.
July 7 is also the day Claude Fable 5 users begin facing their first real cost decision. Anthropic has announced that starting July 7, Fable 5 will shift to a credit-based system for usage beyond plan quotas, ending its initial free-access period.
OpenAI is not just launching a model. It‘s capturing the users Anthropic is about to lose. Fable 5’s safety classifier is already frustrating users — some have reported it blocks basic queries (like the word “human”) due to false positives in its cybersecurity classifier. The window is open.
This is the same playbook Anthropic used when it launched Fable 5 on June 9 — the day after GPT-5.5‘s free tier ended and many users were looking for an alternative. OpenAI is returning the favor.
July 7 vs July 17: The Model War’s Two Fronts
If July 7 is the first date to watch, July 17 is the second. That‘s when Gemini 3.5 Pro is expected to launch.
Three models. Two weeks. One window.
Company | Model | Date | Positioning |
|---|---|---|---|
OpenAI | GPT-5.6 (Sol/Terra/Luna) | July 7 | Price/agentic play |
Gemini 3.5 Pro | July 17 | Context/frontend play | |
Anthropic | Fable 5 | — | Stuck in safety limbo |
Anthropic is the only major player without a new release. Fable 5’s safety classifier is so restrictive that users are reporting it blocks even harmless queries. The model‘s degraded experience is pushing users to consider alternatives. OpenAI and Google are both competing for the same users.
OpenAI’s three-tier pricing strategy covers the full spectrum: Sol at $5/$30 for premium workloads, Terra at $2.5/$15 for everyday heavy lifting, and Luna at $1/$6 for low-cost tasks. Luna‘s pricing directly challenges the open-source cost floor — and Sol undercuts Fable 5 at half the price.
P.S. July 7 matters for two reasons: the Fable 5 quota expires, and OpenAI is positioned to take its users. The model names are already on GitHub. The rollout is already in motion. A week from today, the model war enters its next phase — and Google, OpenAI, and Anthropic each have different reasons to be worried. All three can’t win. July will decide which one loses.
