Strengths
- A very cheap Contributor mode ($0.10 input, $0.20 output) for experiments and prototypes where you do not mind Meta training on the data.
- Good at finding things in long texts: on MRCR at 512K to 1M tokens it scores 98.1 in Meta's table, against 73.8 for GPT-5.6 Sol.
- On the DeepSWE v1.1 coding test it scores 75.4 in Meta's table, slightly ahead of Claude Opus 5 (74.0) and GPT-5.6 Sol (73.0).
- Much cheaper per task than the flagship: $1.60 versus $4.52 for Muse Wonder in Artificial Analysis testing.
Weaknesses
- Version 1.3 does not fully support audio, and answers may be worse. For audio, Meta recommends version 1.2 or Muse Voice Transcribe.
- It trails Claude Opus 5 on professional office work even in Meta's own table: GDPVal-AA v2 1754 versus 1824.
- Verbose: on one Artificial Analysis task it writes about 60K tokens.
- The cheap Contributor mode means Meta may use your data for training. It also has lower limits and no top thinking level.
Best for
- coding agents on a budget
- long documents up to 1 million tokens
- tight tool loops and helper agents
- inputs with images and video
Not ideal for
- understanding audio
- confidential data in Contributor mode
How to use it in Europe
Through the Meta Model API you can use it in standard mode, where Meta does not train on your data, or in the cheap Contributor mode, where it may. Developers also get it in a Muse Code subscription. The Muse Spark line powers the Meta AI assistant in WhatsApp, Instagram and on meta.ai. Meta does not publish an official country list, though, and does not mention any option to process data in the EU.
Price compared with similar models
Benchmark results
| Benchmark | Score | Source |
|---|---|---|
| Artificial Analysis Intelligence Index v4.3 (max) | 48 | artificialanalysis.ai ↗, October 2026 |
| DeepSWE v1.1 (max, vendor-reported) | 75.4 | dev.meta.ai ↗ (vendor-reported), September 2026 |
| Terminal-Bench 2.1 (max, vendor-reported) | 88.8 | dev.meta.ai ↗ (vendor-reported), September 2026 |
| MRCR 512K-1M long context | 98.1 | dev.meta.ai ↗ (vendor-reported), September 2026 |
| GDPVal-AA v2 | 1754 | dev.meta.ai ↗ (vendor-reported), September 2026 |
Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.
More models from this vendor
- Muse Wonder: Meta's strongest model for long-running agents and coding, still released as a preview.
- Muse Spark Open: Meta's large open model that a company can run on its own servers, so data never leaves its infrastructure.
- Muse Glimmer 30B: Meta's small open model under Apache 2.0 that runs on a single graphics card, even offline.
- Llama 4 Maverick: The larger model in the newest Llama generation, from April 2025. Cheap through providers, but its license excludes people and companies in the EU.
- Llama 4 Scout: The smaller Llama 4 from April 2025, very cheap through providers. Its license grants no rights to people or companies in the EU.
Similar models from other vendors
- Claude Sonnet 5.5 (Anthropic): Claude's balanced model for everyday work and coding. It is faster than Opus 5.5 and costs half as much.
- Command A+ (Cohere): An open business model from Canada's Cohere that officially supports Slovak and Czech and runs on your own server.
- Gemini 3.8 Flash (Google): Google's fast, lower-cost model for everyday work, coding and document analysis. You can try it in the API for free.
- Mistral Medium 3.5 (Mistral AI): Mistral's mid-size open-weight model and the default in the Vibe app. Good for everyday work and coding with data kept in the EU.
- Nemotron 3.5 Super VL (NVIDIA): NVIDIA's open model for business AI agents that understands text, images and video and is very fast and cheap to run.
- GPT-6.1 Sol (OpenAI): OpenAI's balanced model, close to GPT-6 Astra in quality at a fifth of the price. A solid pick for business apps.