Thinkymachines ships Inkling 2 Mini: a 26B-A3B looped transformer that runs on laptops
simonguozirui · x · 2026-11-21
Thinkymachines released Inkling 2 Mini after complaints that Inkling 2 Small (594B) was too big. It's a looped transformer with just 26B total params (26B-A3B, 24 layers, layers 6–17 looped 3x) that runs on a laptop, scoring 20.1 on the AA Index — on par with Qwen3.8-27B — for agentic work. SkyRL offers Day-0 RL support with vLLM; its chart shows full fine-tuning, LoRA and MxFP8 achieving similar rewards on DAPO-Math-17K, with MxFP8 showing lower rollout/trainer logprob divergence.
Related event: Thinking Machines releases open-weights 26B looped model Inkling 2 Mini(23 posts)→
More from Models
- METR Time Horizons 2.0 returns: 3.6-month doubling time, GPT-6 Astra hits 40.3 hours — scaling01 · 2026-11-21
- Google Search AI leaks its chain-of-thought and internal formatting tags to users — ConsistentMap5775 · 2026-11-21
- SenseNova Loop 9B on a single RTX 4080: 256k context for 16GB VRAM GPUs — roofkid · 2026-11-21
- Google's subscription page leaks unannounced "Hydrogen" model in AI Plus tier — giffmana · 2026-11-21
- xAI staff responds to Grok bot computer failures, escalates support ticket — mattyp · 2026-11-21
- Looped transformers: more looping boosts reasoning but with diminishing returns past 3 passes and fewer CoT tokens — nrehiew_ · 2026-11-21