Send your first DeepSeek request
- Create a free TokenReply account and create an API key (walkthrough).
- Use the OpenAI-compatible base URL
https://api.tokenreply.com/v1. - Start with
deepseek-v4-flash, then change only the model ID to try a thinking or Pro model.
curl https://api.tokenreply.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $TOKENREPLY_API_KEY" \
-d '{
"model": "deepseek-v4-flash",
"messages": [{"role": "user", "content": "Explain this function and suggest one improvement."}]
}'The OpenAI Python and JavaScript SDKs use the same request. Set base_url or baseURL to https://api.tokenreply.com/v1; examples are in using your API key.
Which DeepSeek model should you choose?
| Family | Good for | Trade-off |
|---|---|---|
| Flash | Fast chat, extraction, routine coding and high volume | Less deliberate on multi-step problems |
| Thinking | Maths, planning and debugging | Slower and uses more output tokens |
| Pro | Quality-first work and harder coding tasks | Higher price; check the live row |
DeepSeek models in the live catalog
Read from the live catalog when this page opens.
- $0.5 in · $2.15 out per 1M tokens
deepseek-r1Paid - $0.5 in · $2.15 out per 1M tokens
deepseek-r1-0528Paid - $0.3 in · $1.2 out per 1M tokens
deepseek-v3-0324Paid - $0.21 in · $0.79 out per 1M tokens
deepseek-v3.1Paid - $0.27 in · $1 out per 1M tokens
- $0.15 in · $0.75 out per 1M tokens
- $0.078 in · $0.114 out per 1M tokens
deepseek-v3.2Models of the week - $0.078 in · $0.114 out per 1M tokens
deepseek-v3.2-thinkingModels of the week - $0.22 in · $0.66 out per 1M tokens
deepseek-v4-flashModels of the week - $0.22 in · $0.66 out per 1M tokens
- $0.22 in · $0.66 out per 1M tokens
deepseek-v4-flash-thinkingModels of the week - $0.44 in · $1.32 out per 1M tokens
- $0.66 in · $1.98 out per 1M tokens
deepseek-v4-proPaid - $0.66 in · $1.98 out per 1M tokens
- $0.66 in · $1.98 out per 1M tokens
- $0.22 in · $0.66 out per 1M tokens
- $0.22 in · $0.66 out per 1M tokens
Troubleshooting
| Symptom | What it usually means | What to do |
|---|---|---|
| 401 Unauthorized | The key is incomplete, revoked, or was pasted with spaces. | Copy the key again and send it as Authorization: Bearer sk-.... |
| 403 Forbidden | The model is not included in your account tier or in the key's model list. | Pick a Free or Weekly Featured model, or review plans and tiers. |
| 429 Too Many Requests | A rate limit or a daily allowance was reached. | Wait before retrying, lower concurrency, or switch to another available model. |
No upstream channel is available for this model | The model is listed, but no route can serve it right now. | Check status, then try another DeepSeek model. |
Frequently asked questions
Is there a free DeepSeek model?
Only when the catalog marks one as Free. Weekly Featured DeepSeek models include a daily call allowance on Default and Trial accounts; the table above shows the current access for each model.
Do I need a DeepSeek SDK?
No. Use the OpenAI SDK with the TokenReply base URL and an exact DeepSeek model ID.