Thinking Machines Lab releases looped open-weight model Inkling 2 Mini: 26B total, 5.4B active, runs on a Mac in ~20GB
Announced November 20, 2026
Verified (fact)
Thinking Machines Lab released Inkling 2 Mini, an open-weight model presented on the Hugging Face blog and in the Ollama v0.40.17 release. Of roughly 26B total parameters, 5.4B are active per token. A looped architecture repeats the middle layers for each token (three times, according to community summaries), increasing per-token compute relative to the model's size. It can run locally, for example on a Mac, with about 20GB of memory, and supports text and image input, tool calling, and thinking. The same Ollama release also added deferred tool loading for Claude Code (loading MCP tools only when needed) and concurrent request handling on Apple silicon.
Needs fact-check (claim)
79.2% on SWE-Bench Verified, per the Ollama release notes. According to community summaries, among open-weight models with fewer than 100 billion total parameters it leads public evaluations with 48.2% on DeepSWE v1.1 and 46.5% on Vals Index v2.2, while multi-turn conversation and following instructions with complex constraints are cited as relative weaknesses.
Why it matters (interpretation)
Repeating layers to raise reasoning quality without adding parameters, aimed at agentic coding at a size that runs locally, could become a new direction in the competition among small open models.
Sources
- Official Welcome Inkling 2 Mini, Cutting-edge Performance for your Local Setup Hugging Face
- Official v0.40.17 Ollama
- Official v0.40.17 Ollama
- Community Inkling 2 Mini - 노트북에서 실행하는 반복형 트랜스포머 모델 GeekNews
Story thread
- 2026-11-21 Thinking Machines Lab releases looped open-weight model Inkling 2 Mini: 26B total, 5.4B active, runs on a Mac in ~20GB
- 2026-11-21 Inkling 2 Mini follow-up: Ollama v0.40.17 runs it out of the box ('79.2% on SWE-Bench Verified'), adds concurrent requests on Apple silicon
More about Thinking Machines Lab
-
Inkling 2 Mini follow-up: Ollama v0.40.17 runs it out of the box ('79.2% on SWE-Bench Verified'), adds concurrent requests on Apple silicon
Ollama's v0.40.17 release (tagged rc4) supports Thinking Machines Lab's Inkling 2 Mini. It can be run with a single command, `ollama run...
-
Inkling 2 Small follow-up: officially added to Hugging Face Transformers v5.23.0, with 48 MoE layers routing to 8 of 384 experts each
Hugging Face released Transformers v5.23.0, adding Thinking Machines Lab's 'Inkling 2 Small' to its newly supported models. According to...
-
Inkling 2 Small follow-up: Thinking Machines Lab plans to release weights for enterprise customization (report)
According to AI Times, Thinking Machines Lab, led by Mira Murati, launched its second-generation model 'Inkling 2 Small' on its official...
-
Thinking Machines Lab releases 'Inkling 2 Small', a 594B-total, 17B-active MoE with image input and long autonomous tasks (community reports)
According to GeekNews, Thinking Machines Lab has released 'Inkling 2 Small', its second-generation MoE model. It has 594 billion total...