Leaks Suggest Anthropic and OpenAI Are Testing Unreleased AI Models
Social media reports indicate Anthropic may be routing traffic to an unreleased Fable 5.5, while OpenAI's GPT 6.2 Astra appears in usage data, though neither company has confirmed the releases.
Key takeaways
- Users report that Claude Opus 5.5 requests are being routed to an unreleased model potentially named Fable 5.5.
- OpenAI's GPT 6.2 Astra model identifier has appeared in Codex usage data for specific user accounts.
- Neither Anthropic nor OpenAI has officially announced these new model versions, with evidence consisting mostly of removed code snippets and user observations.
Unverified reports circulating on social media suggest that major artificial intelligence developers Anthropic and OpenAI are quietly testing upcoming model versions. According to Decrypt, users have observed technical indicators pointing to the deployment of Claude Fable 5.5 and GPT 6.2 Astra, although neither company has provided official confirmation. The speculation has intensified in the lead-up to Anthropic's initial public offering, with community members analyzing code changes and output quality shifts to predict release timelines.
Evidence of Unreleased Anthropic Models
The rumors regarding Anthropic began in early October when users posted screenshots indicating that outputs from Claude Opus 5.5 appeared superior to typical results. Pseudonymous user Chetaslua noted on X that Opus 5.5 requests were being routed to a new model, which could be either Fable 5.5 or a new Opus variant. This routing process involves sending user queries to a different backend model without changing the visible interface label.
Further technical evidence emerged on November 1, when a label for "Mythos 5.5" was briefly visible in the hidden source code of Anthropic’s official website. Kim Isenberg, a prominent figure in the AI community, shared a screenshot of this code snippet, crediting an anonymous tipster for the discovery. Additionally, an entry for "claude-fable-5-5" appeared in the LiteLLM open-source catalog on October 13, though it was later removed. The catalog entry initially copied specifications from Fable 5.1, with notes indicating that no official price or capability limits were established for the new version.
OpenAI GPT 6.2 Astra Signals
Parallel reports have surfaced regarding OpenAI's next-generation model. According to Decrypt, the identifier "gpt-6.2-astra" appeared in Codex usage data for some accounts on November 1. This suggests that OpenAI may be conducting internal tests or limited external trials of the model before a public launch. Like the Anthropic rumors, these reports rely on user-generated data and interface observations rather than official press releases from the company.
Context and Corporate Timeline
The timing of these leaks coincides with significant corporate milestones for Anthropic. The company filed for an IPO on October 26, reporting revenue of $16.29 billion for the first half of 2026, a substantial increase from the previous year. Anthropic's IPO roadshow is scheduled to begin no earlier than November 10. Historically, Anthropic has managed its release cycles carefully, having previously paused the launch of Fable 5 for 18 days due to U.S. export controls before releasing Fable 5.1 in September. The current speculation suggests a potential acceleration in their product roadmap, with Fable serving as the premium commercial tier and Mythos as a restricted model for select government and enterprise clients.

Discussion 0