Muse Spark 1.3
muse
Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It improves long-horizon agent collaboration, instruction following, and coding efficiency relative to Muse Spark 1.2.
ReasoningTool CallingAttachmentsStructured Output
Context Window
1.0M
Max Output
131K
Temperature
Yes
Open Weights
No
Knowledge Cutoff
N/A
Released
2026-09-02
Last Updated
2026-09-02
Modalities
Input:TextImageVideoPDFAudio
→Output:Text
Benchmarks
| Benchmark | Score | Metric | Harness |
|---|---|---|---|
| GDPval-AA | 1754 | Elo | Stirrup |
| JobBench | 64.9 | mean rubric score | OpenCode |
| OSWorld | 66.9 | mean partial score | Meta internal evaluation framework |
| OSWorld | 32 | binary completion rate | Meta internal evaluation framework |
| DeepSearchQA | 90.3 | F1 | Meta browser harness |
| Agentic IF Index | 57.8 | score | - |
| AutomationBench | 49.6 | pass@1 | - |
| MRCR | 98.5 | mean sequence-match ratio | - |
| MRCR | 98.1 | mean sequence-match ratio | - |
| DeepSWE | 75.4 | pass@1 | mini-swe-agent |
| SWE-Atlas Codebase QnA | 59.4 | pass@1 | mini-swe-agent |
| Terminal-Bench | 88.8 | pass@1 | Muse Code |
Available Providers (5)
| Provider | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | Reasoning | Status |
|---|---|---|---|---|---|---|
 NanoGPT | Rp 22.351 | Rp 75.991 | Rp 2.683 | — | effort | — |
 LLM Gateway | Rp 22.351 | Rp 75.991 | Rp 2.683 | — | effort | — |
 Kilo Gateway | Rp 22.351 | Rp 75.991 | Rp 2.683 | — | effort | — |
 OpenRouter | Rp 22.351 | Rp 75.991 | Rp 2.683 | — | effort | — |
 Vercel AI Gateway | Rp 22.351 | Rp 75.991 | Rp 2.683 | — | effort | — |