GPT 6 Astra API on Sprelay
Use model ID gpt-6-astra on Sprelay's OpenAI-compatible endpoint. The live model page lists available groups and their prices. Group choice affects both access and cost.
Connection and group selection
| Setting | Value |
|---|---|
| Model ID | gpt-6-astra |
| Client base URL | https://sprelaytoken.com/v1 |
| Chat endpoint | https://sprelaytoken.com/v1/chat/completions |
| Publicly priced groups | ChatGPT Dedicated Pro, ChatGPT High-Speed |
Checked against the public catalog on 11 October 2026. Account access and upstream availability may differ. This is a configuration guide, not a claim of an authenticated inference test or a performance benchmark.
Compare dated group prices
The live page showed the following example rates in USD per 1 million tokens. Current live prices and usage logs take precedence.
| Group | Input | Output | Cached input | Cache write |
|---|---|---|---|---|
| ChatGPT Dedicated Pro | $0.90 | $4.50 | $0.09 | $1.125 |
| ChatGPT High-Speed | $1.30 | $6.50 | $0.13 | $1.625 |
For 10,000 uncached input tokens and 2,000 output tokens, the illustrated Dedicated Pro cost is 0.01 × $0.90 + 0.002 × $4.50 = $0.018. The same token counts at the illustrated High-Speed rates cost $0.026.
These examples exclude cache writes, additional calls and other billable usage. Group labels alone do not establish speed or quality. Check billing and measure your actual workload.
Verify a small request
- Create a dedicated key, select an available compatible group and set a small quota.
- Use the first request example with model
gpt-6-astra. - Confirm generated content in
choices, then match the key, model, status and charge in usage logs. - Compare several representative requests before increasing the budget. Do not estimate an entire agent run from a single short response.
Cherry Studio supports custom provider configuration. For Cursor, verify your installed version's model and feature restrictions separately.
Diagnose access or connection errors
A model access error can mean the key uses a different group. A 401 calls for checking the token, expiry and balance. For 404, inspect the actual URL and avoid /v1/v1. Follow API error handling.
Send only redacted diagnostics to support: timestamp, status, model and request ID. Keep the full API key private.