Mercury 3

mercury-3

Diffusion language model for fast reasoning, coding, and tool use

Intelligence: Pending
Context: 1M
Training cutoff: --
Added: 2026-10-20

Providers

One route serves Mercury 3, and every figure below is measured on it. Where the selected window has too little traffic, the figure names what it is based on instead.

Inception only route

mercury-3

Uptime

100.0%

Input / 1M: $0.10
Output / 1M: $0.35
TTFT: 1.07s fixed prompt · n=3
Output rate: 328.6 tok/s fixed prompt · n=3

1M context 131,072 max output cache $0.01 read / $0.10 write per 1M Retention Yes, period not published Training Yes Provider terms, reviewed 2026-10-19warming · 30m

Prompt caching Automatic reuse Eligible Cache key Ignored Caller breakpoints None advertised

TTFT is the median time to the first visible answer token over real ModelRelay requests, with a fixed prompt standing in for models that have no traffic yet. Speed ranks by it. Intelligence is the Artificial Analysis Intelligence Index.