Meta launches Muse Spark 1.3 with coding and agentic capabilities
Meta released Muse Spark 1.3 on September 2, 2026, which the company describes as its largest improvement yet in coding and agentic work. The model scored 88.8 on Terminal-Bench 2.1, matching GPT-5.6 while using fewer tool calls and tokens than its predecessor.
Detected & updated continuously · Source: Nebula
Story subjects
Sources
@tldrnewsletter
Meta released Muse Spark 1.3, which it says is its largest improvement yet in coding and agentic work. It scored 88.8 on Terminal-Bench 2.1, matching GPT-5.6, while making 20% fewer tool calls and using 25% fewer tokens than Muse Spark 1.2. Mark Zuckerberg said open weights
@MSBIntel
Meta launches Muse Spark 1.3 with agentic workflows and coding capabilities, advancing AI agents beyond conversation. https://t.co/hvIsGBgxb7
@robertojirusta
Meta launches Muse Spark 1.3 for coding https://t.co/Y7T5kqebA1 baha breaking news
@markfenner
Damn, I fear Meta is cooking. Muse Spark 1.3 is rolling out today with what Zuckerberg calls Meta’s biggest jump yet in coding and agentic work. On Meta’s numbers, it leads this comparison on long-context retrieval, DeepSWE, and codebase Q&A, while tying GPT-5.6 Sol on
@MTorygreen
Muse Spark 1.3 is getting uncomfortably close to the expensive models on the stuff agents actually do. On Meta’s own evals it’s basically neck-and-neck with Opus 5 on professional tool use and computer use, ahead of GPT-5.6 Sol on several coding tests, and it’s doing this while https://t.co/PDtYeAmpL9
@jiahanjimliu
$META: Huge Release, Muse Spark 1.3 (max) surpasses Gemini 3.8 (High), Sol-5.6 (max), Kimi K3 on benchmarks We are seeing the model layer getting increasingly tight as models converge on performance. If this continues, the every model provider will fight to buy compute capacity