Thinking Machines releases Inkling 2 Small multimodal MoE model
Understand the story
On November 6, 2026, Latent Space published its AI news briefing covering November 4–5, 2026, headlined by Thinking Machines releasing the multimodal MoE model Inkling 2 Small. The model has 17B activated parameters out of 594B total (17A-594B), with pretraining completed in 68 hours. The briefing also aggregated industry updates including the METR report. Only this one report exists so far; no further details or follow-up developments are available.
AI overview of reporting · Updated 1 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Latent Space[AINews] Inkling 2 Small: 17A-594B trained in 68 hours on 12T tokens, beating Muse Spark
Latent Space published its AI news briefing covering November 4-5, 2026, headlined by Thinking Machines' Inkling 2 Small (17A-594B, pretrained in 68 hours), along with the METR report and other industry updates.
Story heat trend
Not enough continuous observations to show a trend yet.