We celebrate each new model as a major step forward - but the inference layer has also improved dramatically in 2026.
Median agentic cache rate is up to 89% from the mid 60s to start the year.
Better cache = lower cost on the same model
OpenRouter says median agentic cache rate rose to 89% in 2026
AIOpenRouter says the median agentic cache rate rose from the mid-60s at the start of 2026 to 89%. It says better caching lowers cost on the same model.
Source details
OpenRouter · x.comPublished · added here
Original title: We celebrate each new model as a major step forward - but the inference layer has also improved dramatically in 2026.