Announcement
Hugging Face: Thinking Machines Lab releases Inkling 2 Mini, with AI layers on repeat
Hugging Face says the local-focused AI model runs its 12 middle layers three times per token.
TLDR
Thinking Machines Lab has released the open-weights Inkling 2 Mini, Techmeme reports. The “Mini” still has billions of parameters: Hugging Face lists roughly 26B total and 3B active per token. Hugging Face also says it accepts text and images, supports a 1M-token context window, and works with Transformers, llama.cpp and MLX.
Combined views
3.2K
3 Sources, first seen ago

Comments