Meta Releases Muse Spark 1.3 Contributor, a Low-Cost Multimodal Reasoning Model With 1M Context Window

Meta has released Muse Spark 1.3 Contributor, a multimodal reasoning model positioned as the budget tier of its Muse Spark 1.3 lineup. The model is now listed on OpenRouter with a 1 million token context window and pricing of $0.10 per 1M input tokens and $0.20 per 1M output tokens.

What Meta Claims

According to Meta, Muse Spark 1.3 Contributor is designed for "experimentation, learning, and early-stage agentic, multi-agent, and coding workflows." The company states the model can track information across extended tasks, work through conflicting inputs, and request clarification or confirmation when needed. These capability claims have not been independently verified against standard benchmarks, and no benchmark scores were disclosed in Meta's release materials.

The listing notes that prompts and outputs sent to this model may be used by Meta to improve its products — a data usage policy worth flagging for developers with privacy-sensitive workloads.

Specifications

Confirmed technical details from the OpenRouter listing:

• Context window: 1,000,000 tokens
• Input pricing: $0.10 per 1M tokens
• Output pricing: $0.20 per 1M tokens
• Cache read pricing: $0.002 per 1M tokens
• Modality: Multimodal (text, image, and audio input/output)
• Availability: Listed via Meta as a provider on OpenRouter, served across all locations

Meta has flagged a known limitation at launch: audio understanding is "currently not fully supported," and response quality may degrade for requests that include audio content. No parameter count, training data cutoff, or standard benchmark results (e.g., MMLU, HumanEval) were published alongside this release.

OpenRouter's uptime tracking shows the model has been live for approximately 3 days, with insufficient data yet to report throughput, latency, or uptime statistics. There is also no usage data yet from applications built on the model.

What This Means

Muse Spark 1.3 Contributor appears to be a stripped-down, low-cost variant of a larger Muse Spark 1.3 model family, aimed at developers who want to prototype agentic or coding workflows without paying premium rates. At $0.10/$0.20 per 1M tokens, it undercuts most flagship multimodal models on price, though the unresolved audio support and absence of published benchmarks make it difficult to assess how it stacks up against competitors like GPT-4o mini or Gemini Flash on actual task performance.

The "Contributor" naming and the explicit note that prompts may be used to improve Meta's products suggest this tier functions partly as a data-collection vehicle — users get cheap access, Meta gets training signal. Teams evaluating this model for production should treat Meta's capability claims as unverified until independent benchmarks appear, and should factor in the data usage terms before routing sensitive workloads through it. The 1M token context window is notable at this price point, but real-world reliability — especially for multimodal and audio tasks — remains unproven given the model's limited track record so far.