Estimate the USD cost of a set of calls

Estimate the USD cost of a set of calls

Usage

gr_estimate_cost(model, input_tokens, output_tokens = 0)

Arguments

model

Model id.

input_tokens, output_tokens

Token counts (scalars or vectors; vectors are summed).

Value

A single numeric USD figure, or NA_real_ when the model has no pricing in the registry. Pricing is seeded from the registry's as_of snapshot. Treat it as an estimate, and use gr_register_model() to correct it.

See also

gr_model_info(), gr_register_model(), gr_trace_summary()

Other cost and token functions: gr_budget(), gr_count_tokens(), gr_model_info(), gr_model_limits(), gr_models(), gr_register_model(), gr_set_tokenizer(), gr_tokenizer(), gr_truncate_tokens()

Examples

gr_estimate_cost("gpt-4o", input_tokens = 120000, output_tokens = 4000)
#> [1] 0.34
gr_estimate_cost("a-model-with-no-pricing", 1000, 100)
#> Warning: Model 'a-model-with-no-pricing' is not in the registry. Falling back to a conservative 128000-token context with a 4096-token output reserve. Register the real limits with gr_register_model('a-model-with-no-pricing', context_window = ..., max_output = ...) so budgets and cost estimates are correct.
#> [1] NA