According to Beating Monitor, just three weeks after the release of GPT-5.5, its successor, GPT-5.6, has already been accessed by external developers. Multiple developers successfully invoked the unreleased gpt-5.6 model within the Codex environment using OAuth authentication via ChatGPT Pro. Probe tests revealed a context window of 1.5M tokens, approximately a 43% increase over GPT-5.5 API’s 1.05M tokens. The earliest traces of GPT-5.6 appeared on April 28. Developer Haider, while reviewing Codex routing logs, noticed that the vast majority of calls targeted gpt-5.5, but one mapping explicitly listed gpt-5.6. He later revised his assessment, suggesting it was more likely a canary test or a bug, as the entry quickly disappeared. However, changes have become evident this week: developers noted that last week, specifying gpt-5.6 returned the error “model is not supported,” but this week, using Pro’s OAuth authentication, the request succeeded without issue. More critically, the context window has surged: while GPT-5.5’s API context is 1.05M tokens and drops to just 400K tokens via Codex OAuth, GPT-5.6 probes directly reached 1.5M tokens—an increase of nearly 1.5 times. Real-world tests in OpenCode confirmed that the model responds normally with inputs exceeding 900K tokens and accepts requests beyond 1.05M tokens. The model self-identifies as running on openai/gpt-5.6 during conversations, supports an xhigh reasoning level, and offers a fast mode with noticeably improved speed. Today, blogger Leo stated that GPT-5.6 development is now fully underway; the first checkpoints began internal testing in recent days, with an expected release next month. He also revealed two internal codenames: ember-alpha and beacon-alpha. Haider analyzed OpenAI’s iteration pace: from annual updates to biannual, quarterly, and now bi-monthly cycles—now shortened to 30–45 days—leading him to estimate GPT-5.6’s release around early June. He further predicts GPT-5.6 will surpass GPT-5.5 on several benchmarks where the latter still lags behind Mythos, arguing that GPT-5.5 is already very close and additional reinforcement learning will widen the gap, given OpenAI’s stronger RL loops in coding, mathematics, and scientific research. On Polymarket, the probability of GPT-5.6 being released before June 30 is currently around 85%.
GPT-5.6 detected in Codex with 1.5M token context, expected June release
MarsBitShare
Summary
On-chain data shows that GPT-5.6 has been detected in Codex with a 1.5M token context, a 43% increase from GPT-5.5. Developers using ChatGPT Pro’s OAuth have successfully invoked the model, confirming its stable operation. On-chain analysis indicates the model was first observed on April 28 and now runs on openai/gpt-5.6 with fast mode support. Internal testing is ongoing, with a June launch anticipated. Polymarket assigns an 85% probability to a June 30 release.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.