Inkling 2 Mini: what it is, is it free, can it run on a Mac?
Thinking Machines' Inkling 2 Mini is a free Apache 2.0 open model (26.4B, 3.4B active) that runs on a Mac with 32GB of memory. Specs, scores, prices.
Checked and updated 21 November 2026 at 03:20 UTC

Key points
- Inkling 2 Mini is a small open-weight AI model from Thinking Machines Lab, released November 20, 2026. It reads text and images, writes text, and is aimed at coding and agent tasks.
- The weights are free under Apache 2.0. The download is about 20GB and runs on an Apple silicon Mac with 32GB or more memory, a DGX Spark, or a GPU with about 32GB of VRAM.
- On Tinker it costs $0.15 per million input tokens and $0.50 per million output tokens, half of Inkling 2 Small. It is weaker at multi-turn chat and fact recall.
Inkling 2 Mini is a free, open-weight AI model from Thinking Machines Lab that is small enough to run on a laptop. Released on November 20, 2026 under the Apache 2.0 license, it reads text and images, writes text, and is built for coding and agent work. You need a Mac with at least 32GB of memory, or a similar GPU, to run it yourself.
It is the little sibling of Inkling 2 Small, the bigger model the company opened up earlier in November. Here is what Mini can do, what it costs and how to try it.
Inkling 2 Mini at a glance
| Inkling 2 Mini | |
|---|---|
| Maker | Thinking Machines Lab |
| Released | November 20, 2026 |
| License | Apache 2.0 (free, commercial use allowed) |
| Size | 26.4B parameters, 3.4B active per token (5.4B counting the loops) |
| Input / output | Text and images in, text out |
| Context window | Up to 1 million tokens |
| Download | One checkpoint of about 20GB (NVFP4 and MXFP8) |
| Tinker price | $0.15 input, $0.50 output per million tokens |
What is special about it
Inkling 2 Mini is a "looped" model. Its middle 12 layers run three times for every word it writes, so it gets more thinking out of a small set of weights. Thinking Machines says this helps it reach the right answer in fewer tokens, especially in coding agents, and that it runs faster than models with similar scores.
The trade-off is knowledge. Looping doesn't let the model store more facts, so its fact recall is in line with other models of its size. The company says it makes up for that by using search and tools well.
How good is it?
Thinking Machines says Mini leads every open-weight model under 100B total parameters on DeepSWE v1.1 (48.2%) and the Vals Index v2.2 (46.5%). Some scores from the model card:
| Benchmark | Inkling 2 Mini | Qwen3.8-27B | Gemma 4 31B |
|---|---|---|---|
| SWE-bench Verified | 79.2% | 77.8% | 46.1% |
| DeepSWE v1.1 | 48.2% | 42.2% | 0.3% |
| Terminal Bench v2.1 | 71.6% | 73.0% | 43.4% |
| BrowseComp | 79.0% | 61.6% | 37.0% |
| MMMU Pro | 74.9% | 76.3% | 73.4% |
These are the company's own numbers, and setups vary by model. It also admits weak spots: it is not the best small model for long multi-turn chats or prompts with many rules (50.0% on MultiChallenge, behind Qwen3.8-27B), and it trails bigger models on the hardest reasoning tests. If you want the Qwen alternative, see our Qwen3.8 27B page.
How to try Inkling 2 Mini
- In the browser: Tinker's playground gives inference-only access.
- On a Mac: Apple silicon with 32GB or more memory. The model card lists Ollama and Exo Labs on Apple silicon, llama.cpp and Unsloth GGUF versions, and the Llama App.
- On a PC or server: a GPU with about 32GB of VRAM, or one NVIDIA DGX Spark, with vLLM, SGLang or Hugging Face Transformers.
- For developers: fine-tune or call it through Tinker, at $0.15 per million input tokens and $0.50 per million output tokens (64K context option; training is $0.75 per million tokens).
Not sure your computer can handle local AI? Our guide to running AI models on your phone or laptop explains the basics.
AI Tools This Week is not affiliated with Thinking Machines Lab.
Frequently asked questions
Is Inkling 2 Mini free?
The model weights are free to download and use, including commercially, under the Apache 2.0 license. Tinker's playground offers inference-only access, and Tinker's paid pricing is $0.15 per million input tokens and $0.50 per million output tokens.
Can I run Inkling 2 Mini on my laptop?
Yes, if it has enough memory. Thinking Machines says the 20GB checkpoint runs on Apple silicon MacBooks, Mac minis and Studios with 32GB or more memory, and lists Ollama, llama.cpp, Llama App, vLLM and Unsloth GGUF versions among supported tools.
What is the difference between Inkling 2 Mini and Inkling 2 Small?
Mini is the small, cheap one: 26.4B total parameters, a 20GB download that fits on one computer, and half the Tinker price ($0.15 vs $0.30 per million input tokens). Small is the bigger, stronger model, and Mini approaches its score only on some tasks such as kernel optimization.
Sources
Related
Apple FastFrames: what it is and is it free to use?
Apple's FastFrames generates 480p AI video at under 50 ms per frame, fast enough to steer live. Free weights, but research-only and it needs big Nvidia GPUs.
FLUX 3 Image Klein: is it free, and can your PC run it?
FLUX 3 Image Klein is Black Forest Labs' 5.6B open image model. It's free for small businesses, runs on 8 GB GPUs, and costs $0.009 per image in the API.
Haru-4B: the open AI voice model that talks and listens at once
Haru-4B from Japan's higurashi is a free, open voice model that listens and speaks at the same time, in 9 languages. What it does and how to run it.