Fable 5.2 first tests: beats Opus 5.5 on coding, price down 20%
量子位 · wechat · 2026-11-10
Early community tests of Fable 5.2 show strong benchmarks: 45.5% above Opus 5.5 on a coding test, 69.9% on SWE-bench-style evaluation (+13.6 pts over its predecessor), up to 30% faster output, 1M-token context, and a 20% price cut ($8/$40 per million input/output tokens).
Highlights from users:
- A Cyber-Call-of-Duty playable game built in 8 hours with zero human feedback; an entire playable Earth map with live weather; an underwater game shipped in 76 minutes for $40
- Pure-code generated short films, a SaaS ad in 10 minutes from one prompt, full piano sonata composition-to-visualization
- A 3D room built in Blender from one prompt (52 min, $5 render cost); Ethan Mollick ran 25k agents to spawn 450+ code-generated 3D worlds; a da Vinci manuscript revived as interactive 3D
Complaints: token burn at max reasoning tiers is steep, and some tasks improve little over Opus 5.5 while costing twice as much.
Related event: Claude Fable 5.2 launch sparks wave of one-shot generation demos(28 posts)→
More from coding & agent
- Anthropic's real bet is MCP connectors embedding HubSpot, Adobe and Atlassian inside Claude — petrusenko_max · 2026-11-11
- ZenML Pro launches always-on AI engineering team to keep agents in production — strickvl · 2026-11-11
- Latent Space tests NaiveAI's 2,100 tok/s coding model: fast, verbose, promising — Latent Space · 2026-11-11
- Anaconda's LLMOps guide: retrieval quality degrades independently of the model — anacondainc · 2026-11-11
- Dev demos agent with 'durable object brain' plus on-demand computer, local or remote — samgoodwin89 · 2026-11-11
- Right-sizing intelligence per role: frontier reasoning at the control plane, adversarial verifiers at the gate — clmt · 2026-11-11