Meta Muse Spark 1.3: Cost-Effective Legal AI Performance

Meta's Muse Spark 1.3 (Max) model shows performance similar to Fable 5 and GPT 5.6 Sol, while being 4 to 8 times cheaper. This comes from internal benchmarking by Vals AI. What's interesting is its strong performance in legal applications. It ranks first on Vals AI's Legal Research Benchmark and second on Harvey's Legal Agent Benchmark. The model also completes tasks quickly, with fewer conversational turns and faster queries. This could mean lower costs and higher output for users. However, the testing did show content refusals on sensitive legal topics, like export controls, in 10 out of 1,327 tasks. The model was evaluated with a 1 million token context window and 131,000 maximum output tokens. Pricing is listed at $1.25 and $4.25 per million tokens for different usage tiers. The bottom line is that these findings could impact pricing for competing models and influence decisions in areas like legal research and contract analysis.