OpenAI says its GPT-5.6 family is now available across ChatGPT, Codex, and the API, with GPT-5.6 Sol positioned as the flagship tier. The company also previewed a Cerebras-powered deployment of Sol that it says can reach up to 750 tokens per second for select customers.
OpenAI announced GPT-5.6 Sol, Terra, and Luna, and said its flagship Sol model will be available in a limited Cerebras-powered preview at up to 750 tokens per second for select customers.
OpenAI’s GPT-5.6 release page says the model family is generally available across ChatGPT, Codex, and the API. The company describes GPT-5.6 Sol as the flagship tier in the family, alongside Terra and Luna variants intended for different performance and cost needs.
OpenAI’s developer model documentation lists GPT-5.6 Sol with a 1,050,000-token context window and up to 128,000 output tokens. The same OpenAI Developers page lists API pricing for GPT-5.6 Sol at $5 per million input tokens and $30 per million output tokens.
Notebookcheck, citing OpenAI’s release materials, also reported that GPT-5.6 became broadly available on July 9 and that OpenAI is bringing GPT-5.6 Sol to Cerebras hardware at up to 750 tokens per second.
In its “Previewing GPT-5.6 Sol” post, OpenAI said GPT-5.6 Sol would launch on Cerebras in July at up to 750 tokens per second for select customers. That figure refers to generation speed, a metric that can materially affect how quickly users receive long answers, code, or multi-step reasoning outputs.
The company’s announcement frames the Cerebras deployment as a preview rather than a universal rollout. OpenAI did not say in the cited materials that every ChatGPT or API user will receive 750-token-per-second performance, and the preview language indicates that access will initially be limited.
The sources also do not provide enough detail to independently compare the total cost of the Cerebras deployment with other AI infrastructure options. OpenAI’s public materials cited here focus on model availability, context length, pricing, and the planned high-speed Sol preview.
For developers and enterprise users, higher token throughput can reduce waiting time for long completions. A model with a large context window and high output limit may be used for tasks such as reviewing large document sets, generating extended code, or producing long analytical responses. In those cases, faster token generation can change the experience even if the underlying model quality is unchanged.
However, token speed is only one part of model selection. OpenAI’s published pricing means heavy output use of GPT-5.6 Sol can become expensive relative to smaller or lower-priced models. Organizations evaluating Sol will likely compare latency, output quality, context length, and token cost against their workload requirements.
The confirmed details from OpenAI and corroborating coverage are narrower than some online claims around the release. OpenAI says GPT-5.6 is available across ChatGPT, Codex, and the API; OpenAI Developers lists GPT-5.6 Sol’s context, output, and pricing details; and OpenAI’s preview post says Sol will run on Cerebras at up to 750 tokens per second for select customers.
Claims about total parameter counts, exact infrastructure spending, user conversion economics, or broader business effects are not established by the cited OpenAI and Notebookcheck sources. For now, the news is that OpenAI is pairing its flagship GPT-5.6 Sol tier with a limited high-throughput Cerebras preview while offering the broader GPT-5.6 family through its main products and developer platform.
OpenAI announced GPT 5.6 Sol, Terra, and Luna, and said its flagship Sol model will be available in a limited Cerebras powered preview at up to 750 tokens per second for select customers.
GPT 5.6 expands across OpenAI products OpenAI’s GPT 5.6 release page says the model family is generally available across ChatGPT, Codex, and the API.
The company describes GPT 5.6 Sol as the flagship tier in the family, alongside Terra and Luna variants intended for different performance and cost needs.
Continue reading
