AI modelsNovember 06, 2026
-26
🧊

Inkling 2 Small: 594B params, nearly open

A new language model called Inkling 2 Small is introduced, featuring 594 billion parameters in a mixture-of-experts design with vision capabilities. According to analyses it matches Gemini's tier and would become the top open model outside China once its weights are released under Apache 2.0.

#MoE#Multimodal#Open Weights#Tinker#Reasoning
Inkling 2 Small: 594B Parameter, fast offen
Share Article
šŸ”„ What happened Inkling 2 Small just dropped: a Mixture-of-Experts model with 594B total parameters but only 17B active. It handles image inputs and variable reasoning effort, scoring 31 on the Artificial Analysis Intelligence Index — a clear jump over the previous generation. šŸ’” Why it matters The weights are slated for Apache 2.0 release next week after external safety testing. Until then it's Tinker-only with closed weights. On agentic coding and long-horizon tasks it trades blows with models many times its active size. ⚔ Our take A 17B-active model claiming frontier-adjacent performance is a real flex. But the benchmark isn't the story — the Apache 2.0 drop is. If "next week" slips, the goodwill evaporates fast.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions — when in doubt, read the linked original source.