OpenAI is preparing to ship the GPT-5.6 family as soon as next week, according to TestingCatalog, which first spotted early signals of the release on June 18. The rollout is expected to include the base model alongside Mini and Pro variants, though OpenAI’s history of staggering ChatGPT, API, and Pro access across multiple days means the simultaneous launch is not guaranteed.
The most significant reported capability change is the context window. GPT-5.5 shipped in April with a one-million-token limit. GPT-5.6 is rumored to push that to 1.5 million tokens, a 50 percent increase that would meaningfully change how developers structure long-horizon coding tasks. TestingCatalog also reports improvements to Codex response latency, which has been a friction point for developers using OpenAI’s agentic coding infrastructure.
Early outputs from a GPT-5.6 Pro build have surfaced among a small number of Pro subscribers. One user noted on X that comprehension appeared “increased a lot” compared to GPT-5.5 Pro, while also flagging that response times had stretched back to 20 to 40 minutes for complex tasks, a regression toward GPT-5.5 Pro’s earlier behavior. Frontend and web development tasks showed no visible improvement in early testing. Those early impressions come from an unverified user on X and should be read with appropriate skepticism, not as confirmation of the model’s final capabilities.
The competitive framing around this release is difficult to separate from its timing. Anthropic’s Claude Fable 5, a model in the Mythos tier, is currently subject to a US regulatory action that has left its availability uncertain for some users and markets. TestingCatalog notes that OpenAI already undercuts Anthropic by roughly half on per-token pricing, and the GPT-5.6 launch is expected to deepen those cuts. Whether that pricing move is a direct response to the regulatory situation or a pre-planned margin play is not clear from available reporting.
That context matters for operators currently building on Anthropic’s stack. A window has opened in which OpenAI has a capacity and pricing advantage at the high end of its model line, and the company appears positioned to use it. The regulatory picture around Claude Fable 5 is the variable that most determines how long that window stays open.
GPT-5.6 is not the only product in motion. TestingCatalog also reports that GPT-Bidi-1, a next-generation voice model, is expected to ship around the same time. It is described as a bidirectional audio system capable of listening and speaking simultaneously, handling interruptions, and adjusting mid-sentence. Inside ChatGPT it would appear as a separate option alongside the existing Advanced Voice Mode, with High, Medium, and Instant quality tiers. A draggable voice bubble has been spotted in interface previews.
The release announcement has not come from OpenAI directly. All reported specifications come from TestingCatalog’s tracking of developer signals, early subscriber access, and social media posts, and should be treated as reporting on expected capabilities rather than confirmed product specifications.
For teams evaluating whether to migrate long-context workloads from Anthropic to OpenAI, the next two weeks are the clearest comparison window in months: benchmark GPT-5.6 against your current stack before committing to any contract renegotiations.
Reporting by TestingCatalog (Alexey Shabanov), published June 18, 2026, at testingcatalog.com.
