Running 1,000 workflows on the Anthropic API costs between $2.65 and $60.00, depending on the model choice and assuming a volume of 1,500 input and 500 output tokens per run.
14-day free trial. No credit card required.
For a standard workflow utilizing 1,500 input tokens and 500 output tokens, 1,000 runs will cost between $2.65 and $60.00. The widely used Claude 3.5 Sonnet model sits in the middle of this range at $12.00 per 1,000 workflows.
These estimates are based on official per-token rates. Because output tokens are typically priced significantly higher than input tokens—often 5x to 10x more—the total cost per workflow is highly sensitive to the length of the model's response.
Assumption (stated, adjustable): each workflow run uses 1,500 input tokens and 500 output tokens.
| Model | Input $/1M | Output $/1M | Cost per workflow | Cost per 1,000 workflows |
|---|---|---|---|---|
| Claude API | $0.1 | $5 | $0.0027 | $2.65 |
| Anthropic Claude 3.5 Sonnet (Public Extended Access) | $3 | $15 | $0.0120 | $12 |
Cost per workflow = (input tokens x input price + output tokens x output price) ÷ 1,000,000, using verified official per-token prices.
14-day free trial. No credit card required.
Selection depends on the complexity of the task versus the unit cost. While Haiku 4.5 offers the lowest entry point at $2.65 per 1,000 runs, more sophisticated models like Claude 3.5 Sonnet ($12.00) or Claude Opus ($24.00) may be necessary for tasks requiring higher reasoning capabilities. IT and Finance teams should audit the required intelligence level to ensure they aren't overpaying for simple tasks.
Since input and output tokens have different price points, the cost per 1,000 workflows shifts based on the ratio of prompt length to response length. For instance, Claude API (Opus 4.1) charges $15 per 1M input tokens but $75 per 1M output tokens. Managing the verbosity of the model is a direct lever for controlling the total cost of the workflow.
Measuring the success of Anthropic adoption requires looking beyond the API bill to the net return on spend. By comparing the cost per 1,000 workflows against the baseline cost of manual labor or legacy software, NetLift identifies which models are generating a positive margin. This allows finance teams to validate that a $12 or $60 spend on 1,000 workflows is actually reclaiming enough human time to justify the investment.
14-day free trial. No credit card required.
All pricing on this page comes from official vendor pages:
14-day free trial. No credit card required.
About the author
Paige Gilmore · Founder, NetLift
Paige Gilmore is the founder of NetLift, the AI Value Management platform that helps organisations measure the cost, savings and return of AI adoption.
Paige Gilmore on LinkedInFor a team of 100 users, Gemini for Google Workspace costs between $700 and $2,200 per month, or $8,400 to $26,400 per year. The specific total depends on whether you select the Starter, Standard, or Plus plan.
Read ai-costsFor a standard workflow using 1,500 input and 500 output tokens, 1,000 runs cost between $0.90 on GPT-5.6 Luna and $135 on GPT-5.5 Pro.
Read ai-costsMicrosoft does not publish a fixed price per 1,000 messages for Copilot Studio. Instead, the service is billed through $200 monthly credit packs or via a pay-as-you-go model through an Azure subscription.
Read ai-costsIntercom Fin costs $990 per 1,000 resolutions. The pricing is fixed at $0.99 per successful AI outcome, with no seat requirements for the standalone agent.
Read ai-costsFor 100 users, ChatGPT Business costs $2,000 per month on an annual plan ($24,000/year) or $2,500 per month on a monthly plan ($30,000/year). The investment breaks even when the total group saves between 27 and 33 hours of work per month, assuming a $75/hour loaded labor cost.
Read