OpenAI launches GPT‑6, a unified ChatGPT‑Codex‑browser agent as $852B valuation draws

OpenAI rolled out GPT‑6 on Tuesday, merging ChatGPT, Codex and a web‑browser into a single agent, according to a post by Harsh Mathur on X (formerly Twitter) 【source】. The new model features a 2 million‑token context window that works without the degradation seen in earlier large‑window attempts, allowing developers to feed days‑worth of IoT telemetry data in one call 【source】. Mathur says the dual‑tier reasoning architecture automatically switches between “fast” and “slow” thinking modes, effectively pairing a junior developer’s speed with a senior architect’s depth within the same model 【source】.

Pricing remains unchanged from GPT‑5.4 at $2.50 per 1,000 tokens for input and $12 per 1,000 tokens for output, while Mathur claims GPT‑6 delivers a 40 percent boost in coding and reasoning performance 【source】. He frames the move as a “subsidy war” that benefits builders, noting that unified agents will handle straightforward workflows where full context fits in a single request, whereas specialized multi‑agent systems will still be needed for more complex domains 【source】.

The launch coincides with heightened investor scrutiny of OpenAI’s soaring $852 billion valuation, which the Financial Times reported is under pressure as the company shifts strategy 【source】. Reuters noted that investors are questioning whether the valuation is justified given the pivot toward a monolithic model architecture and the broader market’s expectations for sustained growth 【source】. The scrutiny comes as OpenAI seeks to balance rapid capability gains with cost stability, a balance highlighted by the unchanged token pricing despite the model’s expanded abilities 【source】.

Analysts, per the FT report, warn that the valuation surge could be vulnerable if the unified‑agent approach fails to deliver expected enterprise revenue or if competition from specialized AI providers intensifies 【source】. OpenAI’s leadership has not publicly responded to the investor concerns, but the GPT‑6 rollout signals a bet on the power of a single, highly capable model to streamline development and reduce orchestration overhead 【source】.