Pyyan / Open-Weight Models / Inkling 2 Small

T

Inkling 2 Small

A 594B multimodal model with 17B active, trained natively in 4-bit NVFP4, under Apache 2.0.

currentThinking Machines Lab ↗Verified 2026-11-14

594B

total, 17B active

Specifications

Context1Msource · as of 2026-11-14
Input ($/Mtok)Weights onlysource · as of 2026-11-14
Output ($/Mtok)Weights onlysource · as of 2026-11-14
LicenceApache 2.0source · as of 2026-11-14
Max outputNot publishedsource · as of 2026-11-14 · medium confidence
Parameters594B total, 17B activesource · as of 2026-11-14
Released5 Nov 2026, weights 12 Novsource · as of 2026-11-14
Model IDthinkingmachines/Inkling-2-Smallsource · as of 2026-11-14

What matters

  • Weights published on Hugging Face on 12 November 2026, a week after a hosted-only launch on Tinker on 5 November
  • Takes text and images, writes text, with a context window of up to 1M tokens
  • Pretrained, midtrained and post-trained in NVFP4. The main checkpoint needs at least 500GB of GPU memory, such as 2 B300s or 4 H200s
  • An MXFP8 checkpoint is also published and needs at least 800GB. The lab says it scores only 1 to 2 points higher
  • Local serving recipes for SGLang, vLLM, TokenSpeed and Unsloth. The benchmark scores on the card are the lab's own

Best for self-hosting a large text and image model on a multi-GPU node, or fine-tuning it on Tinker.