Muse Wonder
No published training cutoff
muse-wonder-1.0
$4 in, $20 out per 1M tokens
Muse Spark Open (816B-A42B)
Open-weight sibling of the hosted Muse Spark models; engines: SGLang, vLLM, TensorRT-LLM, Transformers, MLX, llama.cpp, Unsloth. </content> </invoke>
Model idmeta-models/Muse-Spark-Open-816B-A42B
Price$1.25 in, $4.25 out per 1M tokens
Checked15 October 2026
Knows up to January 2026
1,194events after its cutoff
42 historic, 257 major. 257 days of news.
It launched on 15 October 2026, already missing 1,175 of them.
How to call it
5 ways to use Muse Spark Open (816B-A42B), with the model id each provider expects.
| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face (gated, manual approval) | meta-models/Muse-Spark-Open-816B-A42B |
huggingface.co | |
| OpenRouter | meta/muse-spark-open |
openrouter.ai | |
| Baseten, Fireworks AI, Groq, Modal, Nebius, Together AI (hosted APIs named by Meta) | research.meta.ai | ||
| Amazon SageMaker AI, Google Cloud Model Garden, Microsoft Foundry (weights) | research.meta.ai | ||
| Apple MLX (local, 2–4 M5 Ultra Mac Studios) | machinelearning.apple.com github.com |
Pricing
| Input | $1.25 per 1M tokens |
|---|---|
| Output | $4.25 per 1M tokens |
As published: “per 1M tokens (USD) on OpenRouter at launch”.
Source: openrouter.ai, checked 15 October 2026.
Specs
- Input
- Text and images
- Output
- Text
- Context window
- 1,000,000 tokens
- Max output
- Not published
- Open weights
- Yes
- Licence
- muse-model-license (limited commercial terms)
- Training cutoff
- January 2026
- Released
- 15 October 2026
- Status
- Current
- Type
- Reasoning model
What stands out
What its maker marketed at launch, and what people found later, each with its source.
Largest US open-weight model at release
816B total / 42B active natively multimodal MoE with 1M context; Meta calls it ‘the most capable US open-weight model to date’ (own evaluation).
Frontier-scale model on desktop machines
First
Runs in 4-bit NVFP4 (~493 GB) across two to four M5 Ultra Mac Studios linked by Thunderbolt 5 RDMA; about 25 tok/s on two and 37 tok/s on four (Apple ML Research).
Four assistants, one scale
Muse Spark Open (816B-A42B) is added as a fifth row. Each row runs from what the model knows (graphite) through the months between its cutoff and its launch (hatched) to today (pink).
- GPT-6 Astra 261 at launch, 1,114 today
- Claude Opus 5.5 352 at launch, 1,049 today
- Gemini 4 Argon 1,039 at launch, 1,140 today
- Grok 4.7 365 at launch, 1,081 today
- Muse Spark Open (816B-A42B) 1,175 at launch, 1,194 today
Show as a table
| Model | Cutoff | Released | Missing at launch | Missing today |
|---|---|---|---|---|
| GPT-6 Astra | April 2026 | 3 September 2026 | 261 | 1,114 |
| Claude Opus 5.5 | June 2026 | 22 September 2026 | 352 | 1,049 |
| Gemini 4 Argon | March 2026 | 13 October 2026 | 1,039 | 1,140 |
| Grok 4.7 | May 2026 | 21 September 2026 | 365 | 1,081 |
| Muse Spark Open (816B-A42B) | January 2026 | 15 October 2026 | 1,175 | 1,194 |
Notes
First Muse model with open weights. Knowledge cutoff inferred from the chat template default cutoff_date 2026-01-04 on Hugging Face (not stated in prose). Reasoning effort low/medium/high/xhigh (default xhigh). Benchmarks not yet independently reproduced.