Meta Muse Spark 1.1 Tops Blind Agentic Model Benchmark
On X, Garry Tan and Alexandr Wang cast Meta's model as a standout for agent harnesses, while the public evidence remains mostly benchmark screenshots and user claims.
Garry Tan wrote on X that Meta's Muse Spark 1.1, which he said was previously called Hornbill in early access, performed "really good" on his OpenClaw setup. In follow-up posts, Alexandr Wang said on X that Muse Spark 1.1 is "one of the best models" for OpenClaw, HermesAgent and similar harnesses, and another Wang reply called it "one of the best agentic models on the market." The benchmark details circulating with those posts describe a seven-way blind voice-and-judgment evaluation where Muse Spark 1.1 scored 9.00 and finished ahead of rivals including GPT-5.5, but so far the public evidence in view is the X posts and attached image.
Combined views
45.3K
Meta Muse Spark 1.1 Tops Blind Agentic Model Benchmark
On X, Garry Tan and Alexandr Wang cast Meta's model as a standout for agent harnesses, while the public evidence remains mostly benchmark screenshots and user claims.
Garry Tan wrote on X that Meta's Muse Spark 1.1, which he said was previously called Hornbill in early access, performed "really good" on his OpenClaw setup. In follow-up posts, Alexandr Wang said on X that Muse Spark 1.1 is "one of the best models" for OpenClaw, HermesAgent and similar harnesses, and another Wang reply called it "one of the best agentic models on the market." The benchmark details circulating with those posts describe a seven-way blind voice-and-judgment evaluation where Muse Spark 1.1 scored 9.00 and finished ahead of rivals including GPT-5.5, but so far the public evidence in view is the X posts and attached image.