Pricing
A TraeWork-native documentation page for comparing model token rates, platform billing paths, feature multipliers, managed-agent runtime, and billing considerations in one readable view.
Model pricing
Compare model rates, cache hits, and output pricing
for the core TraeWork tiers.
| Model | Input | Cache hit | Output |
|---|---|---|---|
| TraeWork N-500 | $10 / MTok | $1 / MTok | $50 / MTok |
| TraeWork N-500L Limited | $10 / MTok | $1 / MTok | $50 / MTok |
| TraeWork P-480 | $5 / MTok | $0.50 / MTok | $25 / MTok |
| TraeWork R-460 | $3 / MTok | $0.30 / MTok | $15 / MTok |
| TraeWork R-450 | $3 / MTok | $0.30 / MTok | $15 / MTok |
| TraeWork S-450 | $1 / MTok | $0.10 / MTok | $5 / MTok |
Prompt cache writes are 1.25x for 5m and 2x for 1h. Cache hits are 0.1x of base input.
Deprecated and retired model rows are intentionally omitted here to keep the reference scannable. Check the TraeWork reference when maintaining legacy integrations.
TraeWork platform pricing
Review API, routed runtime, regional runtime,
and unit billing options in one place.
TraeWork API
Uses the published token pricing table. Global routing is the default unless a supported data residency parameter is used.
TraeWork Routed Runtime
Pricing is operated by TraeWork. R-series and later distinguish standard and dedicated endpoints, with dedicated routing carrying a premium in the TraeWork reference.
TraeWork Regional Runtime
Pricing is operated by TraeWork. Standard, regional, and dedicated endpoint types are documented separately in the TraeWork reference.
TraeWork Usage Units
Usage is rated by TraeWork, converted into TraeWork Units, then shown through TraeWork workspace billing.
TraeWork Usage Units
See how TWUs convert token usage into
workspace billing and monthly reports.
| Concept | Detail |
|---|---|
| Billing unit | TraeWork Unit, abbreviated as TWU |
| TWU price | $0.01 per TWU |
| Conversion | Token usage is rated in USD, then converted into TWUs for workspace reporting. |
| Metering cadence | Hourly TraeWork metering with monthly invoices. |
| Payment model | Postpaid only; the TraeWork reference notes no prepaid credits for this path. |
| Inference geography | For supported P-series and R-series versions, region_scope: "core" applies a 1.1x multiplier. |
Feature-specific pricing
Scan modifiers for cache, residency, fast mode,
batch jobs, context, and tools.
Prompt caching
- 5m write: 1.25x base input.
- 1h write: 2x base input.
- Cache hit: 0.1x base input.
Data residency
Supported first-party requests with US-only inference carry a 1.1x multiplier across token categories.
Fast mode
TraeWork preview fast mode applies premium input and output rates for supported N-series versions and is not available with Batch API.
Batch processing
Batch API applies a 50% discount to input and output token rates for asynchronous workloads.
Long context
The TraeWork reference states that selected N-series, P-series, and R-series versions include the full 1M token window at standard pricing.
Tool use
Tool use includes normal input and output tokens, tool definition overhead, tool call/result content, and server-side tool charges where applicable.
TraeWork Managed Agents pricing
Understand token plus runtime billing
for TraeWork managed-agent sessions.
Billing dimensions
Worked example
Additional considerations
Use these notes to control cost, capacity,
discounts, and billing setup.
Optimization strategies
- Match model tier to task complexity.
- Use prompt caching for repeated context.
- Use Batch API for non-urgent volume work.
- Watch usage patterns by tokens and tool calls.
Rate limits
Usage tiers affect request capacity. The TraeWork reference lists entry, growth, established, maximum standard, and enterprise paths.
Volume and enterprise
High-volume, enterprise, academic, and research arrangements may use negotiated discounts or custom terms.
Payment basics
Billing is based on monthly usage, paid in USD, with card and invoicing options depending on account type.
Frequently asked questions
Quick answers for token accounting,
cache discounts, and payment setup.
How is token usage calculated?
Tokens are text units processed by a model. The exact count depends on language, content type, and tokenizer version.
Can prompt caching and Batch API discounts stack?
The TraeWork reference says they can be combined, with caching multipliers and batch discounts applied to the relevant token categories.
What payment methods are supported?
Standard accounts can use major credit cards, while enterprise customers can arrange invoicing or negotiated billing terms.