Meta undercuts the AI market with $4.25 model
meta launched muse spark 1.1, a multimodal reasoning model with a 1m token context, priced at $4.25 per million output tokens, signaling a shift toward aggressive pricing and agent-based architectures.
meta released muse spark 1.1, a closed-weight model capable of orchestrating multi-agent systems, handling computer use, and managing a one-million token context window. the model launches alongside the first meta model api at $4.25 per million output tokens, a price point that significantly undercuts recent market leader xai grok 4.5 and establishes a new floor for frontier model costs. the aggressive pricing targets pure-play ai labs like openai and anthropic by squeezing margins between well-funded conglomerates and low-cost alternatives. benchmark claims show muse spark 1.1 leading tests like mcp atlas (88.1) and humanity's last exam (62.1), positioning meta as a top contender in agent performance.