Mythos-class Claude Fable 5 arrives on GitLab Duo Agent Platform

What this post added

This post announces the integration of Claude Fable 5, a Mythos-class model from Anthropic, with the GitLab Duo Agent Platform. Key technical advancements highlighted include: - **Enhanced Goal-Directed Workflows:** Claude Fable 5 can sustain multi-step, goal-directed work with measurably fewer iterations, improving efficiency for tasks like multi-file refactors, incident investigation, and infrastructure-as-code definitions. - **Improved First-Shot Correctness:** The model demonstrates higher accuracy on complex, well-specified problems, reducing the need for back-and-forth interactions. - **Advanced Image and Screenshot Interpretation:** Claude Fable 5 interprets dense technical images and detailed screenshots with significantly higher accuracy and fewer tokens. - **Long-Horizon Autonomy:** The model sustains productive output over extended periods, completing multi-day runs and self-correcting through verification loops. It also dispatches and sustains parallel sub-agents more reliably. - **Reduced Human Oversight Cost:** The operational implication is a reduction in human oversight cost per agent run, allowing teams to tackle harder problems with greater confidence. - **Increased Bug-Finding Recall:** Claude Fable 5 shows noticeably higher bug-finding recall, improving outage triage, root cause analysis, and repository history investigation. - **Deeper Code Reasoning:** The model produces more actionable code review comments by reasoning through code paths with greater depth and identifying edge cases more consistently. - **Availability:** Claude Fable 5 is available across all tiers and deployment models through GitLab's AI Gateway, with specific availability details for GitLab.com and self-managed instances.