~/OPEN SOURCE /meta-open-sources-muse-spark-1-2-amid-llama-4-behemoth-release

Meta Open-Sources Muse Spark 1.2 Amid Llama 4 Behemoth Release Delays

Meta has open-sourced Muse Spark 1.2, a coding-focused model update featuring a 1M token context window and improved debugging capabilities. Meanwhile, the public release of Meta's larger foundation model, Llama 4 Behemoth, remains delayed. Muse Spark 1.2 provides developers with a highly cost-effective open-source tool for complex coding workflows. However, the delay of Llama 4 Behemoth highlights Meta's current strategy of prioritizing smaller, distilled models over its massive teacher models. Muse Spark 1.2 operates at a cost-effective rate of $0.40 per task on the Meta Model API. In contrast, Llama 4 Behemoth is still in training and serves primarily as an internal teacher model to improve smaller models like Scout and Maverick through codistillation.

## BACKGROUND

Meta's Llama series represents a family of open-weights large language models, with the Llama 4 generation introducing native multimodal capabilities. Muse Spark is a specialized model line from Meta optimized specifically for developer workflows, code generation, and codebase understanding.

## REFERENCES

## KEYWORDS

#Open Source AI#LLMs#Reddit Discussion

$ subscribe --daily

Meta Open-Sources Muse Spark 1.2 Amid Llama 4 Behemoth Release Delays | Daily News