Meta's Muse Spark 1.3 Matches OpenAI's GPT-5.6-Sol at Over 90% Lower Training Cost
Latent Space reports that Meta's Muse Spark 1.3 model has achieved performance parity with OpenAI's flagship GPT-5.6-Sol model. Meta reportedly achieved this milestone with a training cost reduction of over 90%, positioning Meta Superintelligence among top frontier AI labs. Achieving top-tier AI reasoning capabilities at a fraction of traditional training costs directly challenges the business models of frontier labs like OpenAI. This cost efficiency could significantly lower the barrier to deploying state-of-the-art AI and intensify competition across the AI ecosystem. Muse Spark 1.3 is a multimodal reasoning model engineered for long-horizon coding and agentic workflows, offering a 1,048,576-token context window. By comparison, OpenAI's GPT-5.6-Sol serves as the baseline flagship model for high-complexity technical problem solving.
## BACKGROUND
Frontier AI development traditionally requires massive capital expenditures to train high-capability large language models. OpenAI introduced the GPT-5.6 series, with GPT-5.6-Sol as its top variant for enterprise reasoning, coding, and complex problem-solving. Meta has pushed for high-efficiency architectures and competitive model releases to challenge existing proprietary leaders.