As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/m/inkling-2-mini/ # Inkling 2 Mini Thinking Machines Lab, Inkling 2 family. Status: current. Type: reasoning model. Released: 2026-11-20. Training cutoff: not published. ## How to call it | Provider | Model id | Endpoint | Docs | |---|---|---|---| | Hugging Face | | | https://huggingface.co/thinkingmachines/Inkling-2-Mini | | Tinker (Thinking Machines) | | | https://thinkingmachines.ai/news/inkling-2-mini | | Local (Ollama, llama.cpp GGUF, vLLM, SGLang) | | | https://huggingface.co/blog/inkling2mini | ## Pricing - Price not published. - Last checked: 2026-11-21 ## Specs - Input: text - Output: text - Context window: not published - Max output: not published - Open weights: yes - Licence: apache-2.0 ## What it has not seen - No published training cutoff. Counted from its release (2026-11-20): at least 9 AI events since (2 major, 0 historic), as of 2026-11-22. - Everything since October 2026: https://postcutoff.com/since/2026-10/ ## What stands out - Looped transformer for local agentic work: 26.4B total / 3.4B active MoE; the two middle six-layer blocks run three times per token (48 layer applications); ~20 GB in NVFP4, runs on a laptop; DeepSWE v1.1 48.2%, AA Index v5.0 20.1. Source: https://thinkingmachines.ai/news/inkling-2-mini ## Notes Benchmarks are the lab's own. Context length, Tinker pricing and modality beyond text not stated on the launch page; HF licence field apache-2.0. Thinking Machines Lab's laptop-sized Inkling 2 model (Nov 20, 2026). Sources: https://thinkingmachines.ai/news/inkling-2-mini and https://huggingface.co/thinkingmachines/Inkling-2-Mini Model page: https://postcutoff.com/m/inkling-2-mini/