Fable 5.2 benchmarks leak: 49.1% FrontierCode, robotics jumps to 64.9%, 89.3% computer use
ChrisGPT · x · 2026-11-10
Fable 5.2 benchmark scores surfaced, showing massive gains over 5.1:
- FrontierCode 2.0 (agentic coding): 49.1% vs 32.4%; FrontierSWE v2 at 69.9%
- Remote Labor Index: 41.3% vs 17.9%
- Computer Use (OSWorld 2.1): 89.3%
- Robotics RoboDojo Sim: 64.9% vs 27.6% — steepest jump
- Mythos 5.2 leads cybersecurity (76.3% CyScenarioBench, 89.6% CVE Rediscovery), protein design (85.1%), and Terminal-Bench-Science (74.6%)
- AA-Briefcase 66.1%, Elo 1785, closing in on Opus 5.5's 1800
The poster argues Anthropic could have called it 5.5, notes heavy focus on robotics and bio, and says chip design benchmarks are coming.
Related event: Claude Fable 5.2 Launches, Leading Opus 5.5 and GPT-6 Across Benchmarks(5 posts)→
More from Models
- Leaked METR docs: Claude Mythos 5.2 is Model 3 Early, may fully automate Anthropic's AI R&D — davidmanheim · 2026-11-10
- A year ago it was "OpenAI smarter, Anthropic better at coding" — now Anthropic seems to win at everything — zack_overflow · 2026-11-10
- Melbourne lab i14labs launches Sugar Glider: O(1) inference architectures tested up to ~1T params — kimmonismus · 2026-11-10
- Claude Fable 5.2 Tops Intelligence Index at 65, Price Cut 20% but Cost per Task 2.6x Opus 5.5 — ArtificialAnlys · 2026-11-10
- Fable 5.2 pricing leaks: $8/1M input, $40/1M output, 1M-token context — Nickleback69420 · 2026-11-10
- Zvi on new Fable benchmarks: if it were really scary they'd have called it 5.5 — TheZvi · 2026-11-10