GPT-5.6 Sol to Appear in Codex Soon
The new GPT-5.6 Sol model will soon be available in Codex, promising impressive processing speeds. A limited test on Cerebras is also planned for July.
GPT-5.6 Sol to Appear in Codex Soon
The model will be in Codex, confirmed by employees. Along with its smaller counterparts Terra and Luna, it is rumored to be released as soon as tomorrow or the day after. We are preparing to vibe-code at full throttle.
What’s More Interesting?
The most interesting part is that a limited test of GPT-5.6 Sol will be released on Cerebras in July. They promise a speed of (attention!) up to 750 tokens per second.
This will not be some separate specialized model. It is the same powerful Sol model that will run twice as fast as Gemini 3.5 Flash.
Model Size
According to some estimates, the model size ranges from 2 to 4 trillion parameters (~150B active). At such a size, fitting the model on Cerebras is impossible without engineering tricks. Therefore, it is assumed that some custom KV-cache scheme has been implemented in the new lineup.
This is something we are truly looking forward to.
Why it matters
AnalysisThis news is significant as the new GPT-5.6 Sol model promises substantial improvements in data processing speed, which could impact the development of AI technologies. It also showcases engineering innovations that may change approaches to utilizing large models.
Discuss in community
Share your questions and insights with developers