Fishing in troubled waters? GPT-5.6's three major models fully revealed, scheduled for July 7th

Author: ASI Revelation; Source: Xin Zhi Yuan

GPT-5.6, to be released next week?

Just yesterday, netizens were excited to discover: In the underlying code of the Codex application, three sub-model identifiers—GPT-5.6 Sol, Terra, and Luna—were found.

Even more anticipated, a brand-new "speed dial" function also appeared in the code.

This suggests that users can freely adjust between speed and quality based on their needs, which will undoubtedly give everyone an unprecedented sense of control.

According to leaks, OpenAI has set a hard internal deadline: The target window for GPT-5.6's release is directly set for next Tuesday (July 7) to July 9.

Why July 7? Because that day happens to be the vacuum period when the specific quota plan for Claude Fable 5 expires.

This is a precisely calculated business hunt down to the hour.

Recently, Anthropic has driven countless developers crazy with a series of messy operations, while Google Gemini 3.5 Pro was forced into an emergency "reboot." OpenAI is seizing this opportunity to prepare for a bottom-fishing move!

Deconstructing the Codex Code: Sol, Terra, Luna Are Coming

"To be honest, OpenAI just quietly stuffed model names into dead code as if we wouldn't notice," joked one netizen.

Since the limited release of GPT-5.6, geeks have been closely monitoring every frontend update from OpenAI.

Finally, in a recent Codex application merge, someone found traces of GPT-5.6.

Another netizen shared a short video. Although the backend interface currently restricts successful model calls, the front-end pop-up window clearly shows the styles of the three models and the new "speed selector."

Moreover, the code also faintly reveals the words "Sol Ultra." Industry insiders speculate that Sol Ultra will be the ultimate trump card directly competing against top-tier flagship models, with performance on par with Fable 5 but at a much more affordable price.

Besides these three models, the code also reveals a key piece of information: The highly anticipated "real-time voice support" is still under development and likely won't launch directly next week.

Leaked Real Test Results: GPT-5.6 vs Fable 5

Although most people haven't used it yet, a few players with internal testing permissions have shared comparison evaluations of GPT-5.6 in real engineering environments.

The result boils down to four words: a crushing blow.

Round One: Extreme Tug-of-War Between Efficiency and Understanding

Overseas tech blogger Shivam shared his experience using GPT-5.6-terra and Fable 5 to solve the same complex technical prompt.

Fable-5 started under a 100% 5-hour session limit. This model kept "Thinking" wildly in the background, burning through 21% of the quota limit. Finally, the response it gave was a series of cross-questions, asking him to reconfirm the technical details to be solved.

For the same task, GPT-5.6-terra consumed only 13% of the quota, with an astonishingly fast response speed.

It didn't waste words; instead, it directly and efficiently listed several different methods and architectural paths to solve the problem, and quickly began execution.

Shivam said bluntly: When using Fable, I was constantly thinking about whether it would suddenly downgrade to Opus 4.8; GPT-5.6-terra's decisiveness made me extremely comfortable.

Round Two: Blind Test of WebGL Front-End Hardcore Mini-Game

Oracle Director Gilson Melo gave GPT-5.6 High and Fable 5 High a tough test question:

In a single HTML file, build a fully functional browser-based game from scratch (using WebGL or HTML5 Canvas). The game must have real-time rigid body physics, gravity, and user-controlled paddle/spaceship mechanics. Write complete CSS, JS, and HTML, no omitting any logic, must support mouse drag for real-time physics feedback.

This question severely tests the model's ability to handle extreme details, long code without shrinkage, and underlying physics formula calculations.

The two models showed different strategies in their workflows.

Fable 5 High performed impressively, generating the entire game code with extreme confidence in one go.

GPT-5.6 High paused twice very humanly during generation, actively asking the developer to clarify two final key decisions.

Even more impressively, without being asked, it took the liberty of adding sound effects to the game.

The final results showed that GPT-5.6 High scored more solidly in overall game experience, smoothness of physical collisions, and robustness of details.

In summary, both testers believed that GPT-5.6 was superior in efficiency and response style, especially in clarity and speed when handling complex tasks.

Based on this result, it is definitely worth looking forward to GPT-5.6's launch next week.

Precise Timing: OpenAI Poaches Users While Others Are in Trouble

If the model leak was accidental, the release time is definitely a carefully considered strategy.

OpenAI plans to launch GPT-5.6 on July 7, exactly when Claude users lose access to Fable 5.

Recently, Claude has lost many users, and OpenAI saw the opportunity to take them all in.

One insider revealed: "GPT-5.6's usage quota limits will be significantly relaxed, more generous than Fable 5. Stricter safety guardrails are also being gradually introduced, but won't be as aggressive as Fable's to affect normal use."

User dissatisfaction is high, and the timing is perfect for OpenAI to snatch users

In comparison, Anthropic has recently faced a lot of public anger.

Although Claude Fable 5 has just returned, it has already caused strong user dissatisfaction.

Ask a few random questions, and Fable 5 downgrades to Opus 4.8.

Biomedical engineer Derya Unutmaz tried to get Fable 5 to explain the word "human."

Just typing "Explain human," the model thought for a few seconds and then popped up a "Switched to Opus 4.8" card because Fable 5's safety mechanism determined that this message contained content that needed to be intercepted.

More absurdly, semiconductor analyst Dylan Patel asked an extremely simple question: "How many letters 'r' are in the word 'raspberry'?"

This question was also intercepted, and the interface popped up "Chat paused," indicating that Fable 5's safety mechanism intercepts most topics related to cybersecurity or biology.

Additionally, Opus 4.8 has recently had serious hallucination problems, even displaying other people's information in its own conversations.

This cliff-like drop in user experience precisely creates the best window for OpenAI to seize users.

Moreover, GPT-5.6 is also likely to have a cost advantage.

Leaks show that GPT-5.6 Sol will be more than twice as cheap as Fable 5 due to its higher token efficiency. But the key is whether its performance is comparable to Fable 5.

Some predict that Sol Ultra should be on par with Fable 5, while being cheaper than Fable 5. If this prediction comes true, OpenAI will completely beat its competitors in cost-performance ratio.

Developer Reminder: 4 Codex Reset Quotas, Don't Let Them Go to Waste

Finally, for all hardcore developers preparing to return to Codex, here's a "money-saving/pitfall-avoidance" guide.

According to in-depth digging by the Reflection CTO, if you've accumulated 4 rate limit reset quotas in Codex, immediately check your account backend.

OpenAI's official underlying rules show that these reset quotas are only valid for 30 days. If your first quota arrived around June 11 or 12, they will start to expire in batches around July 12!

If you want to know your exact expiration time, you can have Codex call your ChatGPT token to request this backend API: GET

You will receive a JSON response similar to the following:

If GPT-5.6 is truly unlocked next Tuesday, you'll only have 4 to 5 days to use your first reset quota.

Next Tuesday, OpenAI will likely give everyone a brand-new Reset. So, in the next few days, quickly use your old quotas where it counts.

GPT-5.6, see you next week!