After 100B tokens of testing, Berman calls Fable 5.2 the most autonomous model yet
MatthewBerman · x · 2026-11-10
Matthew Berman burned 100B+ tokens early-testing Fable 5.2 and calls it the best model he's used:
- Most autonomous yet: its assumptions matched his intent 99/100 on loosely specified tasks
- Bug-catching genius: found bugs no other model found across his codebases
- Best at 3D: controls Unreal Engine and Blender precisely (Gears of War clone demo)
- Fast: finds direct paths to solutions rather than raw token speed
- Ultracode: finishes massive codebase reviews in 10–20 minutes, but burns quota fast
- Completeness: Excel clone with full feature parity, a simulated Tokyo in Unreal, rarely needs follow-ups
He'll still run Opus 5.5 daily for quota reasons, reserving Fable 5.2 for the hardest planning tasks.
Related event: Berman Calls Fable 5.2 the Strongest Model After 100B-Token Test(2 posts)→
More from coding & agent
- Meta's coding model Muse Ember goes live on OpenRouter for agent sub-task fan-out — alexandr_wang · 2026-11-10
- Meta's Muse Ember coding model is now available in OpenCode — alexandr_wang · 2026-11-10
- OpenRSI scales recursive self-improvement to SGLang-Omni, +7.6%–344% throughput via 14 merged PRs — ChengleiSi · 2026-11-10
- OpenCode cuts memory 41% and OpenClaw starts 3.1x faster with Bun AoT — steipete · 2026-11-10
- Vibe-coded ambient music DAW 'Ambient Town' in a week with Push support — kieranklaassen · 2026-11-10
- GR00T 2's WAM API lets reasoning agents steer robots by pointing in the camera view — yukez · 2026-11-10