Fresh reports suggest OpenAI’s next major language model, GPT‑5, could be released as early as next week. While the company has yet to confirm a date, recent leaks and insider claims point to an August 2025 launch window. This follows earlier statements from OpenAI CEO Sam Altman, who has repeatedly said the model would debut this year as part of a more unified AI experience capable of selecting the right tools automatically based on user prompts.
Menlo Ventures AI investor Deedy Das posted on X that an anonymous source with “reasonable proof” claims GPT‑5 will introduce several significant technical changes. The reported specifications include a context window of up to one million tokens for input and 100,000 tokens for output, support for the Model Context Protocol (MCP) to link with external data sources, and the ability to make multiple tool calls in parallel. Additional features may include dynamic reasoning modes for both short and long queries, an integrated Code Interpreter, and other advanced tools.
The leak also referenced internal codenames allegedly tied to GPT‑5’s variations: “Nectarine” for the standard model, “Lobster” for a smaller version, and “Starfish” for a more lightweight build. Other reports from AI tracking platforms have spotted additional names — including Zenith, Summit, and o3‑alpha — which could represent either higher-tier or experimental versions.
Separate screenshots, shared by X user chatgpt21 and attributed to a Cursor design lead, appear to show a Mac application running a GPT‑5‑alpha model in testing. There are also indications that Microsoft is preparing a “Smart Mode” for Copilot powered by GPT‑5.
Altman, speaking in a recent interview, described testing the model himself and being struck by its ability to deliver precise answers to a complex query he struggled to interpret. While such anecdotes give a glimpse of GPT‑5’s potential, the actual performance improvements over current models remain to be seen until broader public testing begins.
If the leaks prove accurate, GPT‑5 could mark a significant step in AI tool integration, reasoning flexibility, and large-scale context handling. However, as with all pre‑release information, details may change before the official rollout.
