Pyyan / Compare / Inkling 2 Small vs Muse Spark Open

Inkling 2 Small vs Muse Spark Open

2 of 5

Open-Weight Models · verified 14 Nov 2026

TInkling 2 SmallThinking Machines Labcurrent
Muse Spark OpenMetacurrent
3 slots left
SpecificationInkling 2 SmallMuse Spark Open
SummaryA 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.
Context1M1M
Input ($/Mtok)Weights onlyWeights only
Output ($/Mtok)Weights onlyWeights only
LicenceApache 2.0Muse Model License
Max outputNot publishedNot published
Parameters594B total, 17B active816B total, 42B active
Released5 Nov 2026, weights 12 Nov15 Oct 2026
Model IDthinkingmachines/Inkling-2-Smallmeta-models/Muse-Spark-Open-816B-A42B
CategoryOpen-Weight ModelsOpen-Weight Models
OfficialThinking Machines Lab ↗Meta ↗

Highlighted rows are where these differ.

Inkling 2 Small

  • Weights published on Hugging Face on 12 November 2026, a week after a hosted-only launch on Tinker on 5 November
  • Takes text and images, writes text, with a context window of up to 1M tokens
  • Pretrained, midtrained and post-trained in NVFP4. The main checkpoint needs at least 500GB of GPU memory, such as 2 B300s or 4 H200s
  • An MXFP8 checkpoint is also published and needs at least 800GB. The lab says it scores only 1 to 2 points higher
  • Local serving recipes for SGLang, vLLM, TokenSpeed and Unsloth. The benchmark scores on the card are the lab's own

Best for self-hosting a large text and image model on a multi-GPU node, or fine-tuning it on Tinker.

Full spec sheet →

Muse Spark Open

  • Text, image, video and audio in, text out, with a 1M token context
  • Muse Model License, not open source: attribution required, and a separate agreement above $20m of model-service revenue
  • About 1.5TB in BF16, or two Mac Studios at 4-bit

Best for multimodal work you host yourself.

Full spec sheet →