Twelve ways to pay for Opus-class usage, plotted against monthly input-token volume at a steady output-token volume. Flat seat plans are horizontal; metered APIs rise with usage. Edit the assumptions to match usage.
Prices are published list rates as of October 2026 except Harvey, which is the client's quoted contract. Claude Enterprise: $20/seat/mo billed annually plus usage at API rates, 20-seat minimum. Claude Team: $20/seat/mo annual ($25 monthly), 2–150 seats, usage subject to plan limits rather than metered per token. Opus 5.5 lists at $4.00 per 1M input tokens on Anthropic's API and on the global endpoints of Vertex AI, Amazon Bedrock, and Azure AI Foundry; regional, cross-region, or data-zone endpoints carry roughly a 10% premium on each cloud, so the three cloud lines overlap at list price and separate only when you enter a negotiated or regional rate. Harvey: $60,000/yr for 9 seats, one seat required per attorney regardless of usage. OpenAI: ChatGPT Business (formerly Team) lists at $20/seat/mo annual ($25 monthly) for standard seats and $100 ($125 monthly) for premium, 2–200 seats; Claude Team premium seats are likewise $100/seat/mo annual ($125 monthly) with 5× standard usage; ChatGPT Enterprise is custom-priced, and the $60/seat with a 150-seat minimum used here is the figure reported by third parties, not OpenAI; its usage beyond the seat fee is handled through a negotiated credit pool rather than a published per-token rate, so it is drawn flat. GPT-5.6 Sol lists at $5.00 in / $30.00 out per 1M tokens. The chart's x-axis is input tokens; the expected output tokens from the slider (set to one-tenth of input, 50:5, unless moved by hand) are charged at each provider's output rate and added as a fixed monthly amount to every metered line. Annual figures assume the same monthly usage for 12 months. Caching and batch discounts are excluded.