Pricing

THROUGHPUTS bills per token, passed through at the same rate the upstream provider charges. There's no markup on pay-as-you-go. Committed-use plans discount the per-token rate further.

Plans

Per-token billing

You pay for two things, both measured in tokens:

1. Prompt tokens — your input. Every token in the messages array, including system prompts and tool definitions.
2. Completion tokens — the model's output.

For multimodal requests, image inputs are billed as a fixed token equivalent that varies by model (typically 85 tokens for a low-detail image, 765 for high-detail).

The unit is USD per million tokens. A model priced at $2.40 / 1M in, $9.60 / 1M out will cost:

• 1K prompt tokens → $0.0024
• 1K completion tokens → $0.0096

Estimate a request

curl https://api.throughputs.dev/v1/pricing/estimate \
  -H "Authorization: Bearer $THROUGHPUTS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "monthlyInputTokens": 2_000_000,
    "monthlyOutputTokens": 500_000
  }'

{
  plan: "pay-as-you-go";
  monthlyCostUsd: 9.6;        // 2M * $2.40 + 0.5M * $9.60, divided by 1M
  perMillionIn: 2.4;
  perMillionOut: 9.6;
  breakEvenPlan: "bundle";     // would be cheaper on Bundle at this volume
  breakEvenSavingsUsd: 1.92;
}

What you see on every response

Every completion response includes these headers so you can track spend without a separate billing API:

response = client.chat.completions.create(...)

# Headers are on the underlying httpx response:
in_tok = response.system_fingerprint  # SDK wraps headers; use raw for now

Free tier

Some models (typically smaller embedding or chat models) have pricing.free set to true. Requests to those models don't count against your billing, but they are rate-limited harder than paid models.

Startup program

Qualifying startups get 30% off per-token pricing for 12 months. See Startup Program for eligibility and application.