Meta Muse Spark 1.3 Enters the Coding-Model Race

Muse Spark 1.3 is rolling out through Muse Code, API, and gateway access with a 1M-token context window, high reported throughput, and low pricing. Meta claims coding leadership over GPT-5.6 Sol, but independent tests are mixed and the strongest DeepSWE/reasoning result is reportedly unavailable to developers; available variants show long latency and heavy evaluation-token use. Robust repository-level validation, licensing, limits, safety, and production evidence are still missing.