New Claude checkpoints Fable, Fruitcake leak

// 53d agoNEWS

New Claude checkpoints Fable, Fruitcake leak

New internal model checkpoints from Anthropic, labeled "Claude Fable 5" and "Claude Fruitcake EAP," have reportedly been detected in active testing. This development highlights Anthropic's efforts to bridge the capability gap between its public models and its rumored internal powerhouses like Mythos Preview, indicating that new commercial or early-access versions of their AI may be on the horizon.

// ANALYSIS

Anthropic seems to be actively testing new model architectures under unconventional codenames to fill the gap between their public offerings and internal-only models.

–The shift to names like "Fable" and "Fruitcake" suggests a departure from their traditional tier names (Haiku, Sonnet, Opus), possibly representing specialized or experimental versions.
–The "EAP" label strongly implies an Early Access Program, meaning select developers or partners might already have hands-on access.
–"Fable 5" could be the consumer-friendly or safer variant of their high-capability internal models.

// TAGS

anthropicclaudeclaude-fableclaude-fruitcakeai-modelcheckpointsleaks

DISCOVERED

53d ago

2026-06-09

PUBLISHED

53d ago

2026-06-09

RELEVANCE

6/ 10

AUTHOR

ollobrains

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

RESEARCH1h ago

MANTA enables dynamic topology adaptation for multi-agent systems

MANTA (Multi-Agent Network Topology Adaptation) is a research framework that allows multi-agent LLM systems to dynamically reconfigure their communication topologies at inference time. By combining trace auditing with verbal playbooks during execution, it enables agent teams to optimize collaboration efficiency and achieve superior results on complex benchmarks such as PlanCraft.

NEWS1h ago

Sanseviero highlights persistent workspace challenges in Claude Code

Omar Sanseviero highlighted critical pain points in long-lived Claude Code agent team workflows, where closing terminal sessions destroys active workspace context. Additionally, automatic context compaction blurs granular execution details across individual agents, hindering extended multi-agent development sessions.

OPEN SOURCE3h ago

OpenWorker launches open-source autonomous desktop agent

OpenWorker is an open-source, local-first autonomous desktop co-worker that operates across local documents, terminal commands, and over 25 third-party integrations. Built to execute end-to-end workflows such as file generation and application updates, OpenWorker supports scheduled recurring background jobs while enforcing explicit human approval for high-consequence actions.

New Claude checkpoints Fable, Fruitcake leak