Lanes
Show HN: Lanes now runs Claude Fable 5 across parallel agent sessions
Last verified:
What is Lanes?
Claude Fable 5 is Anthropic's first Mythos-class model to reach general availability, and Lanes exposes it as a runnable model inside its parallel AI coding workspace. It is designed for long, unsupervised coding and knowledge tasks, with strong vision, large context (1M tokens), and the ability to sustain complex multi-step work across days without intervention.
In Lanes, Fable 5 can be selected from the model picker (from v0.43.0) per session or pinned per project via .lanes/cli-flags.json. Each session runs in its own Git worktree with persistent file-based notes, allowing the model to build long-term context and resume work across restarts. Users can control how hard the model works via an
Lanes pricing
Pricing model: Freemium
API pricing is $10 per million input tokens and $50 per million output tokens under the model ID claude-fable-5. Through June 22, Fable 5 is included at no extra cost on Claude Pro, Max, Team, and seat-based Enterprise subscriptions when used via the official claude CLI (which Lanes runs). Starting June 23, usage draws from subscription usage credits rather than being free. Compared to earlier Mythos Preview pricing, this is less than half the cost, but output tokens remain the dominant cost factor.
Lanes pros
- First Mythos-class model in general availability
- Built for long-horizon, unsupervised coding work
- 1M token context window for large codebases
- Up to 128k output tokens per request
- Strong vision understanding for diagrams and charts
- Persistent file-based notes per session in Lanes
- Resume restores full session history across restarts
- Parallel sessions across multiple worktrees
- Effort level knob to control model intensity
- Safe fallback to Opus 4.8 instead of refusing requests
- Included free on Pro/Max/Team/Enterprise until June 22
- Lower API cost than previous Mythos Preview tiers
- Proven on massive migrations (50M-line Ruby in a day)
- Better long-game performance than Opus 4.8 in benchmarks
- Designed for repo-wide refactors and multi-issue workflows
Lanes cons
- Token-intensive by design, especially output tokens
- Conservative on execution, sometimes hesitant to act
- Strong planning but can second-guess follow-through
- Fallback to Opus 4.8 on flagged requests (though rare)
- Free inclusion only until June 22, then usage credits
- $50 per million output tokens can add up quickly
- Requires managing effort levels and prompts carefully
- Most benefit comes from long runs, not quick Q&A
Frequently asked questions about Lanes
What is Claude Fable 5?
Claude Fable 5 is Anthropic's first Mythos-class model available to the general public. It is the same underlying model as Mythos 5 but with safety safeguards in front; when a request is flagged, it falls back to Claude Opus 4.8 instead of refusing, touching fewer than 5% of sessions.
How do I use Claude Fable 5 in Lanes?
From Lanes v0.43.0, you can select claude-fable-5 from the model picker in Session Settings when launching a session, or pin it for a project in .lanes/cli-flags.json. Lanes runs the official claude CLI in a real terminal, which counts as interactive use for billing purposes.
What is the effort level and why does it matter for Fable 5?
The effort level is a knob that controls how hard the model works on a task. For Fable 5, which plans well but can be hesitant on execution, adjusting effort is more effective than arguing with its caution, especially when the model stalls during implementation.
Is Claude Fable 5 free in Lanes?
Through June 22, Fable 5 is free on Pro, Max, Team, and seat-based Enterprise plans when used via the official claude CLI, which Lanes runs. From June 23 onward, it consumes usage credits from your subscription rather than being free.
What is the main use case for Claude Fable 5?
Fable 5 is designed for long, unsupervised runs and ambitious multi-step tasks like repo-wide migrations, large refactors, and agent-style workflows. Its value is in sustained competence over hours or days, not just sharper short answers.
How does Fable 5 handle unsafe or high-risk requests?
When a classifier flags a request as high-risk, Fable 5 does not refuse; instead it falls back to Claude Opus 4.8 and answers there. Anthropic estimates this happens in fewer than 5% of sessions.
What makes Fable 5 different from previous Claude models?
Fable 5 supports much longer autonomous sessions, has state-of-the-art vision, and leverages persistent file-based notes to maintain context over time. Early tests show it can get three times further in long tasks (like a game) and compress months of engineering work into days.
How do I manage multiple Fable 5 sessions in Lanes?
Each Lanes session runs in its own Git worktree with its own history and notes. You can fan out Fable 5 sessions across different modules or issues, view diffs side by side in Git Changes, and get notified when a session needs human input.
What is the API model ID for Claude Fable 5?
The API model ID is claude-fable-5. Pricing is $10 per million input tokens and $50 per million output tokens, and it is the same model as the one exposed in Lanes after v0.43.0.
Should I use Fable 5 for quick questions or small tasks?
Not primarily. Fable 5 is token-intensive and optimized for long-horizon work, so its strengths show up in complex, multi-step tasks. For quick Q&A, it is more expensive and overkill compared to cheaper or faster models.