LongCat-2.0 gets 60% Mercury Cloud discount

This is a compelling price-to-capability play, though developers should validate latency, tool calling, uptime, and discount duration before moving production workloads.

• –LongCat-2.0 activates roughly 48B parameters while supporting a 1M-token context window
• –Its agent-focused training makes it relevant for repository-scale coding and multi-step automation
• –Lower inference pricing could make long-context experiments more economically practical
• –Open weights provide a self-hosting path when hosted pricing or data policies are unsuitable
• –The discount is temporary, so benchmark total cost against established API providers before committing