In the last 90 days, the cost of premium LLM inference has dropped by 42%. We analyze the data from 210+ models to understand why open-source routers like Groq and Together AI are forcing proprietary labs to slash prices.

The implications for SaaS margins are profound. As inference costs approach zero for smaller models, the moat shifts from the model itself to the orchestration layer that routes prompts intelligently.

The Commoditization of Intelligence

When Llama 3 launched on Groq at $0.05/1M tokens, it sent shockwaves through the industry. Within weeks, OpenAI responded with GPT-4o-mini, and Anthropic introduced Claude Haiku. The race to the bottom is accelerating.

For builders, this is a golden age. For labs, it's an existential crisis. The question isn't whether intelligence will be cheap — it's who captures the value when it is.

Partner Application