local… · AGI Hunt

Qwen 3.8 27B vs Claude Fable 5.2: a $0 local model (65 vs 20 on AA Index) and which jobs it can take over

alex_verem · x · 2026-11-17

Anthropic's Claude Fable 5.2 (Nov 9, $8/$40 per million in/out tokens, 1M context) is the strongest closed model; Alibaba's Apache 2.0 Qwen 3.8 27B (Aug) runs free on your own machine. The article compares them: 65 vs 20 on Artificial Analysis Index v5.0.1, 1M vs 262K (extendable to 1M) tokens read at once, data sent to Anthropic vs staying local, 24GB Mac or 16GB GPU suffices. It also lays out a six-step private benchmark you can finish in an afternoon: list paid AI jobs, write 15 real tasks with pass marks, run them on both models via LM Studio/Ollama, score blind, and migrate the tasks Qwen passes to your computer.

Original post →

More from Models

Models channel →