Pyyan / Compare / Inkling 2 Small vs Muse Spark Open
Open-Weight Models · verified 14 Nov 2026
| Specification | Inkling 2 Small | Muse Spark Open |
|---|---|---|
| Summary | A 594B multimodal model with 17B active, trained natively in 4-bit NVFP4, under Apache 2.0. | The largest American open weights, under a licence with a revenue trigger. |
| Context | 1M | 1M |
| Input ($/Mtok) | Weights only | Weights only |
| Output ($/Mtok) | Weights only | Weights only |
| Licence | Apache 2.0 | Muse Model License |
| Max output | Not published | Not published |
| Parameters | 594B total, 17B active | 816B total, 42B active |
| Released | 5 Nov 2026, weights 12 Nov | 15 Oct 2026 |
| Model ID | thinkingmachines/Inkling-2-Small | meta-models/Muse-Spark-Open-816B-A42B |
| Category | Open-Weight Models | Open-Weight Models |
| Official | Thinking Machines Lab ↗ | Meta ↗ |
Highlighted rows are where these differ.
Best for self-hosting a large text and image model on a multi-GPU node, or fine-tuning it on Tinker.
Best for multimodal work you host yourself.