Pricing (Unified)

All prices USD, verified against vendor pricing pages on the lastVerified date above. Pricing moves fast — treat this page as a snapshot, not a spec. Each row links to the vendor's canonical pricing page.

AI coding tools (subscription)

Anthropic models (API pay-as-you-go)

• Batch API = 50% off sync pricing; 24-hour SLA but typically completes in minutes.
• Cache reads = 0.10× of input; cache writes ~1.25× of input.
• See claude.com/pricing for extended thinking, vision, and third-party platform surcharges.

OpenAI models (API pay-as-you-go)

Pricing for OpenAI models fluctuates more than Anthropic's and a 403-blocked fetch during our verification pass prevents a confident table here. Consult platform.openai.com/pricing for current rates; the ballpark order is:

• Flagship reasoning model: highest
• General-purpose chat: mid
• Small / nano tier: lowest
• Batch API: 50% off on all tiers
• Prompt caching: available on most models

Google (Gemini) models

Vertex AI surcharges may apply for enterprise use. Gemini 2.5 family remains available but is legacy; 3.0/3.1 is current.

Framework / platform pricing

These charge for their own hosted infra, not model access:

Gotchas

• Premium-request quotas on Copilot — agent/tool-using tasks burn the monthly quota fast; inline completions are unlimited on paid tiers.
• Cursor's Pro+/Ultra pricing — $60 and $200 tiers buy more premium-model usage; the $20 Pro tier can hit caps with heavy agent use.
• Claude API vs Claude subscription — Claude Code works with either; the Pro/Max consumer subscriptions include Claude Code usage, the API is pay-as-you-go.
• Batch API ≠ real-time. It's a 24-hour SLA (usually much faster) for non-interactive workloads — not for chat UX.

Further reading

• Cost Optimization — how to reduce what you pay
• Tool Comparison Matrix — pricing inside the full feature comparison
• Models — the matching model-capability matrix