Uses published list prices as of 2026-04: Gemma 4 local $0, GPT-4o-mini input $0.15 / 1M tokens + output $0.60 / 1M tokens, Claude Sonnet-4.5 input $3 / 1M + output $15 / 1M. Assumes mean abstract is 220 tokens in + 110 tokens out + 480 tokens of system prompt.
Usage
llm_benchmark_cost(
backend,
model = NULL,
n_abstracts = 100L,
claims_per_abstract = 5,
tokens_in = 700L,
tokens_out = 110L
)Arguments
- backend
String, one of
"ollama","openai","anthropic".- model
Optional exact model id for documentation.
- n_abstracts
Integer; number of abstracts the extractor ran on.
- claims_per_abstract
Numeric; mean claims per abstract (for per-1k-claim normalisation).
- tokens_in
Integer; mean input tokens per call.
- tokens_out
Integer; mean output tokens per call.