Meta's Muse Spark 1.3 Achieves Frontier-Tier Performance at 70% Lower Cost Than Rivals

Meta released Muse Spark 1.3 on September 2, 2026, its fourth Muse Spark model in five months, with the standard "xhigh" variant tying OpenAI's GPT-5.6 Sol (max) and xAI's Grok 4.6 (high) on the Artificial Analysis Intelligence Index while costing 70% less per task. For broader context, explore our AI News.

Meta's Rapid AI Development Continues with Muse Spark 1.3

The introduction of Muse Spark 1.3 on September 2, 2026, highlights Meta's accelerated pace in AI innovation. This model is the fourth in the Muse Spark series to be released within a five-month period, signaling a concentrated effort to advance its AI capabilities and market position. The consistent updates reflect Meta's commitment to pushing the boundaries of what its AI models can achieve.

Frontier-Tier Performance at Unbeatable Prices

Muse Spark 1.3 demonstrates highly competitive performance, particularly in its cost-efficiency. The standard "xhigh" variant of Muse Spark 1.3 achieved a score of 61 on the Artificial Analysis Intelligence Index. This places it on par with leading models such as OpenAI's GPT-5.6 Sol (max) and xAI's Grok 4.6 (high).

Despite matching the performance of these peers, Muse Spark 1.3 (xhigh) costs approximately $0.55 per Intelligence Index task. This is a substantial reduction compared to its 61-point counterparts, GPT-5.6 Sol (max) and Grok 4.6 (high), which are priced between $0.94 and $0.95 per task. This represents a 70% or greater premium for the competing models, making Muse Spark 1.3 a more economical option for similar performance.

Furthermore, the more powerful "max" variant of Muse Spark 1.3 scored 62 on the Artificial Analysis Intelligence Index. This positions it just behind Anthropic's Claude Fable 5.1 and Claude Opus 5, indicating its strong standing among top-tier AI models.

Advanced Agentic Capabilities and Multimodal Support

Muse Spark 1.3 introduces advanced features designed for complex tasks and agentic functions. The "max" variant achieved a 52% score on the Tau3-Bench Banking evaluation, ranking first among all models assessed. This performance highlights its proficiency in handling intricate, domain-specific challenges.

The model also boasts a 1M-token context window, enabling it to process extensive amounts of information. It supports multimodal inputs, including text, image, and video, which enhances its versatility across various applications. These capabilities underscore Meta's focus on developing AI that can manage sophisticated, multi-faceted operations.

What's Next for Meta's AI Ambitions

Meta has outlined future plans for the Muse Spark series, indicating continued investment in its AI ecosystem. The company has announced intentions to develop even larger models, suggesting a trajectory towards more powerful and capable AI systems. Additionally, a Muse Spark open-weights release is on the roadmap, which could foster broader accessibility and collaborative development within the AI community. This move aligns with Meta's historical approach to open-sourcing significant AI research.

Conclusion

Meta's Muse Spark 1.3 represents a significant advancement in the company's AI offerings, combining frontier-tier performance with a highly competitive cost structure. Its rapid release cycle, strong benchmark scores against rivals like GPT-5.6 Sol, and advanced agentic and multimodal capabilities position it as a notable contender in the evolving AI landscape. With plans for larger models and an open-weights release, Meta continues to signal its long-term commitment to shaping the future of artificial intelligence.

Sources

• Introducing Muse Spark: Scaling Towards Personal...
• [PDF] Muse Spark Safety & Preparedness Report - AI at Meta
• [PDF] Muse Spark Contemplating Safety & Preparedness Report - Meta AI
• Meta debuts the Muse Spark model in a 'ground-up overhaul' of its AI | TechCrunch
• Meta AI app climbs to No. 5 on the App Store after Muse Spark launch