GPT-5.6 just officially launched on July 10, and OpenAI’s next move is already in the works. On July 8, the well-known tech leak account @synthwavedd (leo) dropped a bombshell: GPT-5.6 will be the final version in the 5.x series, with GPT-6 expected to debut within a month—or potentially even by the end of July. AI industry analyst Andrew Curran subsequently independently confirmed some of the information and added more critical details.
🚨 SCOOP(s):
– GPT-5.6 will be the final model in the 5.x series. GPT-6 is slated to launch in about a month, earlier than expected, and possibly even later this month
– GPT-6 will be based on a new, significantly larger pretrain (versus the ~4T 5.5/5.6 ‘Spud’ base)
– There is…— leo 🐾 (@synthwavedd) July 8, 2026
All-New Pretrained Base: Farewell Spud
According to @synthwavedd’s leak, GPT-6 will be built on a brand new, significantly larger pre-training base. Previously, GPT-5.5 and GPT-5.6 used a pre-training foundation with approximately 4 trillion parameters codenamed “Spud.” OpenAI originally intended to extend Spud for GPT-6, but ultimately decided to start fresh. Sources say OpenAI is very excited internally about this new foundation, believing it will better compete with Anthropic’s Fable 5 and the upcoming Fable 5.1.
Andrew Curran He further pointed out that this is “a new, much larger-scale pretraining,” and described GPT-6 as OpenAI’s response to Anthropic’s Mythos. He emphasized that the pace of model releases has been accelerating, and GPT-6 will arrive sooner than expected.
Government Regulation: The Biggest Uncertainty Factor for GPT-6
However, even if the model is ready, whether it can be successfully released to the public is another matter entirely. Andrew Curran made it clear: “If the past few months are any indication, GPT-6 will almost certainly be blocked by the government, at least initially.”
This concern is not unfounded. On June 2, 2026, President Trump signed the “Advancing Artificial Intelligence Innovation and Safety” executive order, requiring AI companies to voluntarily submit their most advanced models to the government for evaluation lasting up to 30 days before release. This order has already had a substantial impact on two leading laboratories within the past month.
Everyone is “going big”: Mythos changed everything
Andrew Curran offered a broader observation in a post: “Mythos changed everything. Everyone is going big.” He mentioned that Elon Musk’s xAI is training a Grok model with 10 trillion parameters. Musk confirmed in April 2026 that xAI is training a 10 trillion parameter model, which is the largest publicly confirmed scale among all laboratories. xAI is also simultaneously training up to 7 models in parallel on the Colossus 2 supercluster. In June 2026, Musk announced that Grok V9 completed training with 1.5 trillion parameters, roughly three times its predecessor, but still a huge gap from the 10 trillion parameter Grok 5 target.
SpaceXAI Colossus 2 now has 7 models in training:
– Imagine V2
– 2 variants of 1T
– 2 variants of 1.5T
– 6T
– 10TSome catching up to do.
— Elon Musk (@elonmusk) April 8, 2026
At Anthropic, Claude Fable 5 just came back online not long ago, and Fable 5.1 has entered the late stages of the release process, expected to launch “in the coming weeks.” Andrew Curran concluded: “Both OpenAI and Anthropic are seeing capabilities improving rapidly, and progress will continue along new trajectories in the months and beyond. Both labs are extremely confident in their internal results, believing they have nothing but open sky above them—no ceiling.”
GPT-5.6 has barely arrived, and GPT-6 is already at the door.
A notable context is that GPT-5.6 was just officially released on July 10, including three sub-models: Sol, Terra, and Luna. In the same week, OpenAI also launched the GPT-Live voice model (GPT-Live-1 and GPT-Live-1 mini), which can both listen and speak, making the conversational experience closer to human interaction. If the insiders’ leaks are accurate, GPT-5.6’s “flagship” status may last less than a month. From GPT-5.5 (codename Spud, released April 2026) to GPT-5.6 (July), and then to the possible GPT-6 (late July to August), OpenAI’s release cadence has compressed to a monthly unit.
Conclusion
GPT-6 rumors paint a picture of intensifying AI competition among major labs, with OpenAI abandoning its existing foundation and restarting pretraining; Anthropic’s Fable 5.1 is gearing up; xAI’s 10-trillion-parameter model is in training, and even Meta is rolling out a new Muse Spark 1.1But the biggest variable isn’t on the technical side, it’s on the policy side: the US government’s review mechanism for frontier models has already left its mark on everyone from Anthropic to OpenAI. Even if GPT-6 is technically ready, whether it can be immediately opened to global users after launch still depends on Washington’s approval.
Source: KOCPC Chinese