Inkling 2 Mini is Thinking Machines Lab's open-weights coding and vision model. MTPLX 2.13 runs it with its own draft head, so it writes several tokens per step and every answer keeps the model's own distribution. Every pack needs MTPLX 2.13 or later.
1. Install MTPLX
The Mac app is the easiest way in: download the DMG, drag MTPLX to Applications and open it. It checks your chip and memory, recommends a model that fits, downloads it and sets up its own engine. For the command line alone:
brew install youssofal/mtplx/mtplx
# or
python3 -m pip install mtplx
Inkling 2 Mini needs MTPLX 2.13 or later. Check with mtplx --version; the app updates itself.
2. Pick the pack for your Mac
MTPLX offers only the packs that fit your Mac, in this order:
| Your Mac | Inkling 2 Mini pack to use | Also fits | Context |
|---|---|---|---|
| 8 GB, M3, M4 or M5 | None fits. MTPLX recommends Qwen 3.5 4B Optimized Speed. | ||
| 16 GB to 18 GB, M3, M4 or M5 | None fits. MTPLX recommends Bonsai 2 27B Optimized Speed. | ||
| 24 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | none | 24,576 tokens |
| 32 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | Optimized Speed, Optimized Quality | 163,840 tokens |
| 36 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | Optimized Speed, Optimized Quality | 245,760 tokens |
| 48 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | Optimized Speed, Optimized Quality | 434,176 tokens |
| 64 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | Optimized Speed, Optimized Quality | 749,568 tokens |
| 96 GB to 512 GB, M3, M4 or M5 | Inkling 2 Mini Bare Speed | Optimized Speed, Optimized Quality | 1,048,576 tokens |
| 8 GB, M1 or M2 | None fits. | ||
| 16 GB to 24 GB, M1 or M2 | None fits. MTPLX recommends Qwen 3.5 9B Optimized Speed FP16. | ||
| 32 GB, M1 or M2 | Inkling 2 Mini Bare Speed FP16 | Optimized Speed FP16, Optimized Quality FP16 | 163,840 tokens |
| 64 GB, M1 or M2 | Inkling 2 Mini Bare Speed FP16 | Optimized Speed FP16, Optimized Quality FP16 | 749,568 tokens |
| 96 GB to 192 GB, M1 or M2 | Inkling 2 Mini Bare Speed FP16 | Optimized Speed FP16, Optimized Quality FP16 | 1,048,576 tokens |
In the app, pick the pack by name. From the terminal, download it once:
mtplx pull Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed
On an M1 or M2, use the FP16 build instead: mtplx pull Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed-FP16.
3. Start it with your client
One command starts the model and opens the client on it, already configured:
mtplx start opencode --model Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed
mtplx start pi --model Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed
mtplx start hermes --model Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed
In the app, the same clients are one click each. The app and the command line share one server, so a client started from the terminal uses the model the app already loaded instead of loading a second copy.
4. Send it screenshots
Every Inkling 2 Mini pack carries the model's whole vision tower, so it reads screenshots, diagrams and photos. Paste an image into the app's chat, or send it from Pi or OpenCode as you normally would. In MTPLX's release check every pack read a screenshot of a failing test and named the test and both counts.
5. Use the local API
mtplx serve --model Youssofal/Inkling-2-Mini-MTPLX-Bare-Speed
This serves the OpenAI API (/v1/chat/completions) and the Anthropic API (/v1/messages)
on http://127.0.0.1:8000, with streaming, tool calls and images, and the Responses API
(/v1/responses). Any client that takes a base URL works: Cline, Cursor, Open WebUI, the openai and anthropic Python
libraries, curl.
What speed to expect
MacBook Pro M5 Max with 128 GB, MTPLX 2.13. Decode speed is the fastest recorded figure for each pack, at the model's own sampler: for Bare Speed, a 128-token answer to a 4,142-token prompt already in the prompt cache, recorded through mtplx serve on a 2.13 test build; for Optimized Speed, a 166-token answer in the 2.13 app; for Optimized Quality, the fastest stretch of an answer in the 2.13 app, its first 32 tokens. macOS thermal pressure was nominal for Bare Speed and Optimized Speed, and moderate when Optimized Quality's answer began. First token and prefill: mtplx serve, temperature 1.0 (the model's own sampler), fans at maximum, macOS thermal pressure nominal, on a new 4,142-token coding prompt, three runs for Bare Speed, one run for Optimized Speed on the release build and two runs for Optimized Quality (the mean where there is more than one), timed at the client: the first token is the time from sending the request to receiving it, and prefill is the prompt's tokens over that time. Long sessions on a hot Mac run slower than the peak.
| Pack | Decode speed, peak | First token, new prompt | Prefill | Download |
|---|---|---|---|---|
| Bare Speed | 128.8 tok/s | 1.69 s | 2,466 tok/s | 16.0 GB |
| Optimized Speed | 123.8 tok/s | 1.73 s | 2,397 tok/s | 17.2 GB |
| Optimized Quality | 103.8 tok/s | 1.84 s | 2,257 tok/s | 20.2 GB |
These are rested-Mac figures: a MacBook Pro that has been working hard for a while gets hot and runs slower, even with its fans at full speed, and chips with less memory bandwidth than the M5 Max run slower. M3 and M4 Macs take a code path without the M5's tensor units: on an M5 Max forced onto that path, Bare Speed generated 106.1 tok/s on a cached conversation and answered a follow-up's first token in 0.06 s.
If something does not work
- The pack will not load: update to MTPLX 2.13 or later (
brew upgrade mtplx,python3 -m pip install -U mtplx, or the app's update). MTPLX 2.12 and earlier cannot load Inkling 2 Mini. - Inkling 2 Mini is not offered: your Mac has less memory than the smallest pack needs (24 GB on M3, M4 and M5; 32 GB on M1 and M2). The app recommends the model that fits instead.
- A long conversation is refused on a 24 GB Mac: Bare Speed holds 24,576 tokens there. Start a new conversation, or let your client compact the old one.
- Check what MTPLX sees:
mtplx inspect Youssofal/Inkling-2-Mini-MTPLX-Bare-Speedreports what MTPLX reads in the pack, its vision tower included;mtplx doctorchecks the install.