Which API format should I use?

Use OpenAI-compatible endpoints for compatible SDK and chat workflows. Use Anthropic Messages for Claude Code and Anthropic-style requests. Confirm support on the model or integration page.

How do I switch between supported models?

Keep the same endpoint and key, then change the model ID in the request. Availability and supported features vary by model.

How are model prices presented?

The pricing page lists current discounted input and output rates for each available model. Discounts vary by model, so use the live page as the source of truth when estimating production cost.

Claude Opus 5.5 API: setup, pricing and cost control

Updated October 17, 2026. Use the wetoken.pro model ID claude-opus-5-5. Verify access and pricing for your account before sending production traffic.

Evaluate on real tasks

Anthropic positions Opus 5.5 for agentic coding and knowledge work. Evaluate a migration, repository review or analysis task against explicit acceptance criteria; benchmarks do not guarantee results for your workload.

Confirm model and permissions

Use claude-opus-5-5, not a display label. Check the live catalog, token group permissions, balance and rates. Inspect errors before retrying; do not silently substitute another model.

Messages API setup

POST JSON to https://wetoken.pro/v1/messages with a server-side x-api-key, anthropic-version: 2023-06-01 and Content-Type: application/json. Set max_tokens to bound each response; it does not cap an agent session's total spend.

Claude Code setup

Set ANTHROPIC_BASE_URL to https://wetoken.pro without /v1 and ANTHROPIC_AUTH_TOKEN to your wetoken key in a trusted terminal. Explicitly select claude-opus-5-5 and check the tool's model display and platform usage log. Keep tokens out of screenshots.

Compare prices on equivalent terms

As checked October 17, 2026, Anthropic's announcement lists standard input/output prices of $4/$20 per million tokens. This is an official reference, not a wetoken quote. Compare cache and fast-mode categories separately and check the live wetoken price page for actual rates.

Why one collection task can use many tokens

Agent planning, page fetching, tool output, retries and summaries may involve many model requests. Repeated history increases input usage even with a short final answer. Measure request count, input/output usage, cache categories and completed items.

Reduce waste without silently changing models

Keep the selected model fixed. Extract and deduplicate pages in code, send relevant excerpts, limit tool output and retries, and stop when acceptance criteria are met. Compare total billed cost per accepted result using the same quality checks; a lower price does not guarantee identical quality.

Anthropic Opus 5.5 announcement and reference pricing