gpt-6-astra- Input
- $2.300
- Output
- $11.500
- Cache Write
- $2.875
- Cache Read
- $0.230
Policy-Based AI Delivery Infrastructure
Unified AI Delivery Gateway
Choose a (model + Delivery Lane) pair. We coordinate eligible delivery paths behind a single API for coding agents, AI applications, and automation workflows.
- Request-level visibility
- Session-aware routing
- Cached-token metering
- Pre-commit failover
Illustrative request
Request facts at a glance.
- Requested model
- gpt-5.6-sol
- Served model
- gpt-5.6-sol
- Delivery Lane
- saver95.1%OFF
- Endpoint
- /v1/responses
- Billable usage
- Input tokens6,843Cache read tokens30,742Output tokens413
- Performance
- TTFT4.4sThroughput39.8 toks/s
- Charged
- $0.002350
- Saved
- $0.045579
Delivery Lanes
Same model. Different delivery trade-offs.
Choose a Delivery Lane, then compare its published price with the official model baseline.
Token pricing
Economy Lane: A lower-cost general-purpose lane that favors savings over delivery consistency. Standard context. Standard processing.
| Model ID | Input | Output | Cache Write | Cache Read | Savings |
|---|---|---|---|---|---|
gpt-6-astra | $2.300 | $11.500 | $2.875 | $0.230 | 77.0% OFF |
gpt-6.1-sol | $0.180 | $0.900 | $0.225 | $0.009 | 91.0% OFF |
gpt-6.1-terra | $0.140 | $0.700 | $0.175 | $0.007 | 86.0% OFF |
gpt-5.6-sol | $0.360 | $1.800 | $0.450 | $0.036 | 91.0% OFF |
gpt-5.6-terra | $0.180 | $1.080 | $0.225 | $0.018 | 91.0% OFF |
gpt-5.6-luna | $0.080 | $0.480 | $0.100 | $0.008 | 60.0% OFF |
codex-auto-review | $0.080 | $0.480 | $0.100 | $0.008 | 60.0% OFF |
gpt-5.5 | $0.500 | $3.000 | - | $0.050 | 90.0% OFF |
gpt-5.4 | $0.275 | $1.650 | - | $0.027 | 89.0% OFF |
gpt-5.4-mini | $0.300 | $1.800 | - | $0.030 | 60.0% OFF |
gpt-6-sol | $0.182 | $0.910 | $0.227 | $0.018 | 90.9% OFF |
gpt-6-luna | $0.040 | $0.200 | $0.050 | $0.004 | 60.0% OFF |
gpt-6.1-sol- Input
- $0.180
- Output
- $0.900
- Cache Write
- $0.225
- Cache Read
- $0.009
gpt-6.1-terra- Input
- $0.140
- Output
- $0.700
- Cache Write
- $0.175
- Cache Read
- $0.007
gpt-5.6-sol- Input
- $0.360
- Output
- $1.800
- Cache Write
- $0.450
- Cache Read
- $0.036
gpt-5.6-terra- Input
- $0.180
- Output
- $1.080
- Cache Write
- $0.225
- Cache Read
- $0.018
gpt-5.6-luna- Input
- $0.080
- Output
- $0.480
- Cache Write
- $0.100
- Cache Read
- $0.008
codex-auto-review- Input
- $0.080
- Output
- $0.480
- Cache Write
- $0.100
- Cache Read
- $0.008
gpt-5.5- Input
- $0.500
- Output
- $3.000
- Cache Write
- -
- Cache Read
- $0.050
gpt-5.4- Input
- $0.275
- Output
- $1.650
- Cache Write
- -
- Cache Read
- $0.027
gpt-5.4-mini- Input
- $0.300
- Output
- $1.800
- Cache Write
- -
- Cache Read
- $0.030
gpt-6-sol- Input
- $0.182
- Output
- $0.910
- Cache Write
- $0.227
- Cache Read
- $0.018
gpt-6-luna- Input
- $0.040
- Output
- $0.200
- Cache Write
- $0.050
- Cache Read
- $0.004
Usage observability
See what every request delivered.
Review requested model, served model, Delivery Lane, token usage, cache activity, latency, throughput, and charge for each request.
Illustrative sample data only. These records are not visitor account data or a production performance guarantee.
Built for coding work
A coding task should not stall because one delivery path slowed down.
Long-running coding sessions magnify latency, cache behavior, and manual recovery. Keep those delivery concerns below the workflow.
How delivery works
One API. Multiple delivery paths.
Your workload chooses the public model and Delivery Lane. The policy coordinates eligible delivery paths while the implementation stays behind the API.
- 01Route selectionThe policy narrows delivery candidates from the selected public model and Delivery Lane.
- 02Session affinitySupported long-running sessions can remain aligned to a delivery path when possible.
- 03Same-model fallbackFallback remains within the selected public model when another candidate is tried.
- 04Pre-commit failoverAnother candidate can be attempted before a response is committed when the mechanism applies.
- 05Health signalsRequest outcomes and observations inform the delivery policy without promising a delivery result.
Start with a workload
Move one workload. Keep the rest unchanged.
Use the documented Base URL and a dedicated workspace API key, choose a Published model, Delivery Lane, and endpoint, then verify the first request.
- Base URL
- https://api.1api.club/v1
- API key
- Create a dedicated workspace API key.
- Delivery Lane
- Choose a Delivery Lane for the workload.
- First request
- Verify the completed request in Usage.
Coding agents
Send the first coding task with a Published model, Delivery Lane, and endpoint, then review the completed request in Usage.
AI applications
Send the first application request with a Published model, Delivery Lane, and endpoint, then review the completed request in Usage.
Automation workflows
Run the first automation workflow with a Published model, Delivery Lane, and endpoint, then review the completed request in Usage.
Use Pricing to choose a Published request path, follow Quickstart for the first request, and inspect protocol scope in API Reference.
Trust and FAQ
Clear records. Explicit boundaries.
Review content handling, request facts, pricing, availability, and failure behavior from the public records that define them.
How is API content handled?
Routed API prompts and responses are not written to 1api's application databases or long-term logs. A private continuation cache may temporarily retain the minimum required conversation content for response continuation and one bounded recovery attempt. Read Trust for the complete content boundary and Privacy for provider-side processing context.
What request facts can I review?
Token counts, request status, timing, latency, and charge details stay auditable. Review the customer-facing record expectations in Trust.
How do prices and cached tokens work?
Compare Lane prices with the official baseline. Where a model's public price evidence includes cache pricing, cache read or cache write is shown as a separate pricing line. Cached-token usage is shown in the request record when it applies. Review the selected model's public price evidence in Pricing.
What happens when a delivery path cannot complete?
Where the documented mechanism applies, another candidate can be attempted before a response is committed. The mechanism is not a guarantee that a fallback will occur. The workload should still handle an unsuccessful request. Read Docs for the request workflow.
Which models and endpoints can I use?
Select a Published model, Delivery Lane, and endpoint before sending a request. Use Pricing for publication coverage and API Reference for protocol scope.
Read Trust and Privacy for the complete record and policy details.
Start with a workload
Start with one workload.
Create an API key, choose a Delivery Lane, and follow Quickstart for your first request.