Text and chat · Thinking Machines Lab
Inkling
The first open model from Thinking Machines Lab: general purpose, reads images and audio, 512K context, private mode.
Private Moderate costWhat sets it apart
Inkling is a 975B mixture-of-experts model from Thinking Machines Lab, released with open weights as a broad, balanced base rather than a benchmark specialist. It takes text, images and audio, reasons on demand and supports tool calling.
- 975B total, 41B active, Apache 2.0
- Text, image and audio input
- 77.6% on SWE-bench Verified, per its maker
Strengths and limits
Where it shines
- Reads images and audio as well as text
- Tool calling and adjustable reasoning, private mode
Think twice when
- Mid-range price; cheaper models cover everyday chat
- No video input here
- 512K token context
- 975B parameters, 41B active
- Open weights (Apache 2.0)
Use cases
Ask about a screenshot and a voice memo
Here is a screenshot of our dashboard and a voice memo from my manager. List what she asks me to change and where on the screen.
It accepts images and audio next to text in one request.
Fix a failing test with tools
Run the failing test in the payments module, read the stack trace, and patch the code so it passes without changing the test.
It reaches 77.6% on SWE-bench Verified, per Thinking Machines Lab.
Against its rivals
- Qwen open modelsQwen open models cost less and cover similar multimodal ground.#23 of 90 on Artificial Analysis Intelligence Index (40 points)
- KimiKimi is a close open rival for agent and coding work.#15 of 90 on Artificial Analysis Intelligence Index (44 points)
- Xiaomi MiMoMiMo scores higher on the Artificial Analysis index at a lower price.#32 of 90 on Artificial Analysis Intelligence Index (38 points)
Version history
Each version, dated, with what it changed over the one before.
- InklingLatest
- First open model from Thinking Machines Lab: 975B total and 41B active parameters, up to 1M context
- Multimodal input and balanced skills across reasoning, coding and tool use
Benchmarks
How to read these scores: what a benchmark tells you.
Artificial Analysis Intelligence Index: #56 of 90 models measured in the catalog. One score averaging about ten independent tests of reasoning, knowledge, math, coding and agent work, all run the same way for every model. Higher is better. Each model is listed at its highest reasoning effort. Two or three points apart is a close call; ten points is a different class of model.
See the numbers
| Model | points |
|---|---|
| 1. Claude Opus 5.5 | 58 |
| 2. Claude Sonnet 5.5 | 56 |
| 3. Claude Fable 5.1 | 53 |
| 4. GPT-6 Astra | 53 |
| 5. Gemini 4 Argon | 53 |
| 6. GPT-6.1 Sol | 52 |
| 7. Muse Wonder | 52 |
| 8. Claude Opus 5 | 51 |
| 9. Claude Fable 5 | 50 |
| 10. GPT-6 Sol | 48 |
| 56. Inkling | 25 |
Announced by the maker
Measured by the maker with its own settings: useful, but not directly comparable with the independent scores above.
| Benchmark | Model | Score | Source |
|---|---|---|---|
| swe-bench-verified | Inkling | 77.6 | thinkingmachines.ai |
The model in this family
| Model | Use | Privacy | Price |
|---|---|---|---|
Inklinginkling | Text | Private | $2.276 per million tokens written$0.00159 a page written |
Specifications
| Specification | Inkling |
|---|---|
| Released | July 15, 2026 |
| Parameters | 975B total, 41B active |
| Weights | Open, Apache 2.0 |
| Accepts | text, images, audio |
| Produces | text |
| Context | 524K tokens |
| Longest answer | 66K tokens |
| Abilities | Reads images, Uses Sub Rosa’s tools, Thinks before answering, Tuned for code |
| Price | $2.276 per million tokens written |
| Privacy | Private |
Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.
Sources
Everything on this page comes from these pages, read on October 16, 2026 or as dated on each score.