BREAKING
1B Model Distilled From Claude Fable 5
Frontier Reasoning, Offline Footprint
0
MB
Q4_K_M size
0
GB
Q8_0 recommended
0
K
context tokens
Quantized Build Sizes
Q4_K_M
657
Q5_K_M
751
Q8_0
1100
F16
2100
Runs Local, But Has Limits
Strengths
Apache-2.0
●
Runs on llama.cpp, Ollama, LM Studio
●
Thinking mode with chain-of-thought
●
V2 build adds tool-calling
Cautions
●
Limited on complex tasks
●
Thinking output can be verbose
●
Independent benchmarks still limited
Compressing Frontier AI Locally
AI NEWS BLITZ
A tiny one-billion-parameter model trained on Claude Fable 5 reasoning is now out in GGUF.