Cost per task
What it costs to get a job completed by a model, including hidden reasoning tokens, retries and tool calls, as opposed to the list price per token.
Independent evaluators such as Artificial Analysis now publish cost per task alongside benchmark scores because models with identical token prices can differ threefold in what they cost to finish the same work. A model that emits fewer output tokens, or declines rather than guesses, is cheaper per task even at the same list price. Budget forecasts built on token prices alone systematically underestimate agent workloads.
Related terms
Active parametersAPIAttentionGPUInference
Get the brief in your inbox
One email every morning: the stories that mattered in AI, each with a one-line "why it matters" for your role. Free, no ads, unsubscribe any time.
Published daily by AI Briefing. We never share your address.