Muse Spark 1.2
meta/muse-spark-1.2
Muse Spark 1.2 is a reasoning model from Meta, designed for complex agentic tasks. It accepts text, images, video, and PDF documents, returns text, and offers a 1M-token context window.
Key facts
- Input price
- $1.25 / 1M tokens
- Output price
- $4.25 / 1M tokens
- Context window
- 1M tokens
- Max output
- 944K tokens
- Accepts
- Text, Images, Files, Video
- Returns
- Text
Example: 1,000 requests with 1,000 input and 500 output tokens each would cost about $3.38 (1M input tokens × $1.25 + 0.5M output tokens × $4.25). Estimate your own usage.
Prices are the provider’s own per-token rates with zero markup, from our price list as of Oct 18, 2026. If a call has extra provider charges (images, audio, web search, cached prompts), you pay the reported cost.
Call it from your code
curl https://api.tokenctr.com/v1/chat/completions \
-H "Authorization: Bearer $TOKENCENTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "meta/muse-spark-1.2", "messages": [{"role": "user", "content": "Hello!"}]}'Works with any OpenAI-compatible SDK: set the base URL to https://api.tokenctr.com/v1. Quickstart
Benchmarks
Artificial Analysis
- Intelligence index
- 39.6
- Coding index
- 72.2
- Agentic index
- 43.2
Design Arena
| Category | Rank | Elo | Win rate |
|---|---|---|---|
| Models: ASCII art | 4 | 1295 | 57.9% |
| Models: Data visualization | 8 | 1337 | 58% |
| Models: Websites | 8 | 1315 | 52.1% |
| Models: SVG | 9 | 1287 | 55.6% |
| Agents: Android native | 10 | 1235 | 53.1% |
| Models: Code | 11 | 1315 | 52% |
| Agents: Web apps | 13 | 1216 | 50.6% |
| Agents: Agentic game dev | 15 | 1179 | 42.7% |
| Agents: Full-stack | 15 | 1210 | 48% |
| Agents: Slides (python-pptx) | 15 | 1198 | 48.6% |
| Models: UI components | 15 | 1302 | 52.1% |
| Models: 3D | 18 | 1302 | 51.8% |
| Models: Game dev | 18 | 1304 | 50.3% |
| Agents: HTML slides | 20 | 1135 | 40% |
| Agents: Mobile apps | 20 | 1174 | 47.5% |
Source: Artificial Analysis indexes and Design Arena results as published in OpenRouter’s model data, fetched Oct 15, 2026. Higher is better. Not measured by TokenCenter. See all rankings.