UniRouteClaude Code API pricing and actual usage costs
Compare current UniRoute Claude API token rates, prompt caching, and a reconciled historical bill. Estimate Claude Code costs without double-counting tokens.
Claude Code API costs depend on the model, the context sent to it, generated output, and prompt caching. A coding task can make several requests, so its cost is not determined by the length of your final answer alone.
When you connect Claude Code to UniRoute with a UniRoute API key, check the selected model's UniRoute rate and reconcile usage with your UniRoute account charge. An Anthropic subscription and a UniRoute balance are separate purchases; topping up one does not add funds to the other.
API billing or a subscription
| Question | Claude subscription | Claude Code through UniRoute API |
|---|---|---|
| Which account pays? | Your Anthropic account under its plan rules | Your UniRoute account balance |
| What should you compare? | Plan price, included access, and usage limits | Model rates, request usage, and actual charges |
| Are balances shared? | No | No |
| What is a useful first check? | Check the current plan and authentication rules | Check model availability and start with a small request |
Use the option that fits your workload. A subscription price alone does not establish whether API access will be cheaper for your project. Gateway authentication and current client behavior are described in the Claude Code gateway documentation.
Model rates and billing units
Compare normal input, output, cache writes, and cache reads separately. Each rate applies to its own usage category; a single “input price” does not describe every billable token.
The table below shows default UniRoute public-route rates in USD per million tokens, without personal account discounts. Its retrieval time and rate snapshot identify the displayed rates. Your key’s permissions and route settings determine which route you can use. Compare the separate Anthropic official reference table, checked against the latest official pricing documentation, and use the matching UniRoute rate when estimating a UniRoute charge. A provider's published list price is a reference, not a promise that this is the amount UniRoute will charge.
Default public-route rates · USD per 1,000,000 tokens. “Standard” and “discounted” describe price bands, not the upstream provider. Where several routes share a band, this table shows one complete lowest-price rate set; your key may use another route. Rates cover ordinary processing in the standard context tier, excluding long-context surcharges and priority processing.
| Model ID | Price band | Input | Output | Cache read | Cache write (5 min) | Cache write (1 hour) | Standard context limit |
|---|---|---|---|---|---|---|---|
| claude-fable-5 | Discounted | $4.2 | $21 | $0.42 | $5.25 | $8.4 | 1,000,000 tokens |
| claude-fable-5 | Standard | $10 | $50 | $1 | $12.5 | $20 | 1,000,000 tokens |
| claude-fable-5-1 | Discounted | $4.2 | $21 | $0.105 | $5.25 | $8.4 | 1,000,000 tokens |
| claude-fable-5-1 | Standard | $10 | $50 | $0.25 | $12.5 | $20 | 1,000,000 tokens |
| claude-haiku-4-5-20251001 | Discounted | $0.42 | $2.1 | $0.042 | $0.525 | $0.84 | 200,000 tokens |
| claude-haiku-4-5-20251001 | Standard | $1 | $5 | $0.1 | $1.25 | $2 | 200,000 tokens |
| claude-opus-4-6 | Discounted | $2.1 | $10.5 | $0.21 | $2.625 | $4.2 | 1,000,000 tokens |
| claude-opus-4-6 | Standard | $5 | $25 | $0.5 | $6.25 | $10 | 1,000,000 tokens |
| claude-opus-4-7 | Discounted | $2.1 | $10.5 | $0.21 | $2.625 | $4.2 | 1,000,000 tokens |
| claude-opus-4-7 | Standard | $5 | $25 | $0.5 | $6.25 | $10 | 1,000,000 tokens |
| claude-opus-4-8 | Discounted | $2.1 | $10.5 | $0.21 | $2.625 | $4.2 | 1,000,000 tokens |
| claude-opus-4-8 | Standard | $5 | $25 | $0.5 | $6.25 | $10 | 1,000,000 tokens |
| claude-opus-5 | Discounted | $2.1 | $10.5 | $0.21 | $2.625 | $4.2 | 1,000,000 tokens |
| claude-opus-5 | Standard | $5 | $25 | $0.5 | $6.25 | $10 | 1,000,000 tokens |
| claude-opus-5-5 | Discounted | $1.68 | $8.4 | $0.084 | $2.1 | $3.36 | 1,000,000 tokens |
| claude-opus-5-5 | Standard | $4 | $20 | $0.2 | $5 | $8 | 1,000,000 tokens |
| claude-sonnet-4-6 | Discounted | $1.26 | $6.3 | $0.126 | $1.575 | $2.52 | 1,000,000 tokens |
| claude-sonnet-4-6 | Standard | $3 | $15 | $0.3 | $3.75 | $6 | 1,000,000 tokens |
| claude-sonnet-5 | Discounted | $0.84 | $4.2 | $0.084 | $1.05 | $1.68 | 1,000,000 tokens |
| claude-sonnet-5 | Standard | $2 | $10 | $0.2 | $2.5 | $4 | 1,000,000 tokens |
| claude-sonnet-5-5 | Discounted | $0.84 | $4.2 | $0.084 | $1.05 | $1.05 | 1,000,000 tokens |
| claude-sonnet-5-5 | Standard | $2 | $10 | $0.2 | $2.5 | $2.5 | 1,000,000 tokens |
Retrieved at . Rate snapshot: sha256:ab7d64bde7dea51c8d21a6063819ca4c8e4601685cfcb4a94f4705bbe645a93b. A snapshot identifies the displayed rate set; it is not a frozen billing contract for a future request. Missing cache rates are not free tokens. Confirm the applicable route and billing record before budgeting.
Anthropic official reference rates
Official first-party global API rates were checked on against Anthropic’s pricing documentation. These reference rates exclude priority processing, fast mode, batch discounts and residency premiums. They are separate from the UniRoute retail rates above; a configured channel rate may differ. Only models present in the current UniRoute public price table appear below.
| Model ID | Prompt tier | Input | Output | Cache read | Cache write (5 min) | Cache write (1 hour) |
|---|---|---|---|---|---|---|
| claude-sonnet-5-5 | Standard | $2 | $10 | $0.1 | $2.5 | $4 |
| claude-opus-5-5 | Standard | $4 | $20 | $0.2 | $5 | $8 |
| claude-fable-5-1 | Standard | $10 | $50 | $0.25 | $12.5 | $20 |
| claude-fable-5 | Standard | $10 | $50 | $1 | $12.5 | $20 |
| claude-opus-5 | Standard | $5 | $25 | $0.5 | $6.25 | $10 |
| claude-opus-4-8 | Standard | $5 | $25 | $0.5 | $6.25 | $10 |
| claude-opus-4-7 | Standard | $5 | $25 | $0.5 | $6.25 | $10 |
| claude-opus-4-6 | Standard | $5 | $25 | $0.5 | $6.25 | $10 |
| claude-sonnet-5 | Standard | $2 | $10 | $0.2 | $2.5 | $4 |
| claude-sonnet-4-6 | Standard | $3 | $15 | $0.3 | $3.75 | $6 |
| claude-haiku-4-5-20251001 | Standard | $1 | $5 | $0.1 | $1.25 | $2 |
For Haiku 5.5, prompt length includes ordinary input, cache writes and cache reads. Requests above 100,000 prompt tokens use the higher rate for all token categories in that request. An official model listing alone does not establish that your UniRoute key can use that model.
View prices and recharge options before creating a budget. Use a model that is currently available to your key and compatible with the Anthropic Messages protocol.
Estimate a request without counting tokens twice
For a token-priced request, multiply each usage category by its matching rate, after dividing the token count by one million. Then add any separately documented billing items. The table covers standard token rates; long-context tiers, priority processing, or other separately priced features need their own applicable rates. Do not apply the standard table to those requests.
- Ordinary input uses the ordinary-input rate.
- Output uses the output rate.
- Cache creation uses the applicable cache-write rate and cache lifetime; separate five-minute and one-hour writes when the record distinguishes them.
- Cache reads use the cache-read rate.
Check how the billing record defines input. If a field includes cached tokens, separate those tokens before applying ordinary-input pricing. Apply the rate and accounting rules recorded for that request, rather than recalculating an old request at today's rate.
A task may read project files, use tools, continue through several turns, or start a new context. Record these requests as a task total when estimating a coding workflow. A one-request sample is not a fixed price for completing a programming task.
A reconciled historical request
On October 8, 2026, Claude Code version 2.1.293 made a short production request through UniRoute using claude-haiku-4-5-20251001. The recorded usage and account ledger were reconciled as follows:
| Item | Recorded value |
|---|---|
| Ordinary input tokens | 535 |
| Output tokens | 90 |
| Cache-write tokens | 10,251 |
| Cache-read tokens | 0 |
| Actual UniRoute charge | $0.00579548 USD |
This historical example included cache creation. It is not a current quote, a general per-request price, or a budget for a complete coding project. The charged amount uses the account ledger's precision.
Reconcile your own bill
After a small request, open UniRoute Console and compare the request time, model, recorded usage, and actual charge. With several calls running at once, a balance change can include unrelated requests. Use the matching records rather than attributing the whole change to the last task you started.
Claude Code can show an estimated session cost. The estimate may differ from your gateway rate; the official client also supports configured pricing in some deployments. For a UniRoute request, use the UniRoute billing record to confirm the amount charged. See Claude Code cost management for the client's usage and estimate behavior.
A disconnected stream does not prove that a request consumed no tokens. Check the record before retrying, and retain a redacted error and request time if you need help. Do not share your API key or project contents.
Keep a useful spending baseline
Start with a narrow task, then compare repeated tasks with similar model settings and context. Record task scope, model, token categories, number of requests, and actual cost. This gives you a workload-specific budget instead of relying on a generic monthly estimate.
To set up your connection, follow the Claude Code setup tutorial in Chinese. You can create an API key using the Chinese guide, then check your balance and recharge before making a paid request. Creating a key or reading the guide does not itself establish usable funds.