Confirm identity and visibility first
Use shape, Base URL, model list, and lightweight chat probe together. Compatible OpenAI protocol alone is not identity.
- Model ID: deepseek/deepseek-chat
- Provider: DeepSeek
- Check dimension: limit and speed probe
Model key route
DeepSeek: DeepSeek V3 | limit and speed probe
DeepSeek: DeepSeek V3 limit probing should cover RPM, TPM, concurrency, max output, first-token latency, and occasional 429/5xx risk.
Provider
DeepSeek
11 models in catalog
Context
163,840
China route
Input price
$0.320
USD / 1M tokens
Output price
$0.890
USD / 1M tokens
Key route summary
Model ID
deepseek/deepseek-chat
Provider
DeepSeek
Check dimension
limit-probe
Read-only check. Detection data burns after 5 minutes.
The result shows observed facts only and never stores the full key.
Then split capability and limits
The result shows observed facts only and never stores the full key.
- Parameter surface: supports frequency_penalty
- Parameter surface: supports max_tokens
- Parameter surface: supports min_p
- Parameter surface: supports presence_penalty
- Parameter surface: supports repetition_penalty
- Context: 163,840
- Input price: $0.320 · Output price: $0.890
Finish with operator action
Do not end at success or failure. Return the next action: fill price, add metadata, change route, monitor balance, or do not list yet.
- Read-only check. Detection data burns after 5 minutes.
- China route
- limit-probe