,

Thinking Machines Releases Inkling 2 Mini, a Looped Open-Weights Model That Runs on a Laptop

By the TechMojaz team

Chip illustration, with the headline: Thinking Machines releases Inkling 2 Mini, a looped open model for laptops

Thinking Machines Lab released Inkling 2 Mini on November 20, 2026: an open-weights model small enough to run on a laptop that loops its middle layers to reason more per token and reach answers with fewer tokens.

The model is a Mixture-of-Experts design with 26.4 billion total parameters and 3.4 billion active per token, or 5.4 billion weight activations once looping is counted. Its 24 layers are split into four blocks of six; the two middle blocks run three times for every token, making it a “looped transformer.”

Most of its weights were trained and are shipped in 4-bit NVFP4 precision, giving a memory footprint of about 20GB. At launch it runs through Ollama on Macs with at least 32GB of RAM, the full weights are on Hugging Face, and it can be used and fine-tuned on the company’s Tinker platform.

The lab says it leads all open-weights models under 100 billion total parameters on benchmarks such as DeepSWE v1.1 (48.2%) and Vals Index v2.2 (46.5%). It also notes limits: the model is not the strongest small model for multi-turn chat, and looping improves reasoning but does not add stored knowledge.

Why it matters

Looping is a different route to better reasoning: instead of adding parameters, the model reuses the same layers for more internal computation. That makes capable coding agents practical on local hardware, with no subscription and no data leaving the device. For universities and startups in the Arab region working with tight budgets or unreliable connectivity, that is a meaningful option, though it still needs a well-equipped machine.

Source: Thinking Machines Lab

By the TechMojaz team

اترك تعليقاً