Codex GPT 5.6 Sol Reduced to 258K Context Window (github.com)

OpenAI's Codex has faced significant backlash following a stealthy reduction of its GPT-5.6 Sol context window from 353,400 effective tokens to just 258,400. This change, which represents a 26.9% decrease, was made without prior notice to paid users and is perceived as a severe regression, as the model still advertises a maximum context window of 1.05 million tokens. Such a drastic reduction undermines the reliability of the tool for developers, especially for tasks that depend on maintaining stable long-term context, like repository-scale analyses and large project implementations. The implications for the AI/ML community are substantial, as Codex's decreased context limits make it less competitive compared to rival platforms that offer larger context capabilities, including Google’s Gemini and xAI's Grok models, which boast 1 million-token windows. This not only frustrates existing users relying on the expected functionality of GPT-5.6 Sol but also raises concerns about trust and transparency in OpenAI's product offerings. Users are calling for clear communication about these limitations and immediate restoration of prior context capacities or a pathway to access the published long-context tier to ensure a competitive edge.