Direct calculator

AiCostCalculator is the static service that powers the trait. Use it directly when you don't have an agent instance, or when you want to compute a cost from raw token counts.

All three entry points return an AiCostResult DTO.

From a laravel/ai response

use Aaix\LaravelAiCosts\Services\AiCostCalculator;

$response = $agent->prompt('Hello');

$cost = AiCostCalculator::fromResponse($response);

The model and provider are read from $response->meta, and the token counts come from $response->usage.

From a laravel/ai Usage object

use Aaix\LaravelAiCosts\Services\AiCostCalculator;

$cost = AiCostCalculator::fromUsage($response->usage, 'claude-sonnet-4-6');

// Or pin the provider explicitly
$cost = AiCostCalculator::fromUsage($response->usage, 'claude-sonnet-4-6', 'anthropic');

Useful when you've stored a Usage object separately from the rest of the response.

From raw token counts

use Aaix\LaravelAiCosts\Services\AiCostCalculator;

$cost = AiCostCalculator::fromTokens(
    inputTokens: 10_000,
    outputTokens: 500,
    model: 'deepseek-chat',
);

Use this for backfilling historical data, building cost estimates, or unit tests. If $provider is omitted it is auto-detected from the model name.

Provider auto-detection

Model prefix	Detected provider
gemini*	gemini
gpt* or o\d… (e.g. o1, o3)	openai
claude*	anthropic
mistral*, codestral*, magistral*	mistral
deepseek*	deepseek
llama*, groq*	groq
anything else	unknown

If the model doesn't match any prefix, pass $provider explicitly so LiteLLM can resolve a provider/model pricing key.