DeepSeek V4.1 Flash 1017 free, with no rate limit
An OpenAI-compatible API (/v1/chat/completions). No requests-per-minute cap, no token quota, no waitlist. Change your base URL and you're running.
Currently in beta: free, with no rate limit, while the beta runs. No card needed.
Paste it in a terminal.Paste it in PowerShell. Asks for your key, lets you pick the tools, installs what is missing, backs up every file it changes. Or pick one of 40 tools
Try it here.
The same model the API serves, streaming from the same servers. Text and images.
Like it? Get a free keyOpen the playground and use it in your own code, without the limits of this demo.
No rate limit on your key.
Most model APIs cap requests and tokens per minute, and only raise the cap on higher spending tiers. Past the cap, calls fail with 429.
Here there is no per-key cap. If the service is busy, your request is queued and retried for up to 30 seconds. If it still can't run, you get a 503 with a Retry-After header.
OpenAI-compatible. Change two lines.
Set the base URL and your API key in any OpenAI SDK or tool, and use the model name deepseek-v4.1-flash-1017. Request and response formats are unchanged.
Works with Claude Code, Codex CLI, OpenCode, DeepSeek Harness, Cline, Roo Code, Kilo Code, VS Code Copilot, Cursor, Zed, Aider, Continue and 30 more tools.
What's included.
Model
deepseek-v4.1-flash-1017, a reasoning model.
Rate limit
None. Not per key, not per minute, not per day. Busy moments go through a 30-second retry queue.
Endpoints
POST /v1/chat/completions and GET /v1/models, in the OpenAI format.
Streaming
Server-sent events, as your SDK expects. Closing the connection stops the generation.
Reasoning
Kept apart: thinking in reasoning_content, the answer in content.
Keys
Up to 25 per account, each with its own usage, revocable one by one.
Usage
Requests, tokens, latency and first-token time per key and per day. CSV export.
Logs
We never store prompts or answers. Only time, token counts, latency and status.
Price
Nothing to pay during the beta, and no card to give.
It runs to whoever kneels to drink. An empty cup holds the most. Thought should flow like water. W. move the cursor through the mist
Free during the beta, with no rate limit.
Input and output tokens cost $0 while the beta runs. After it: $0.005 per million input tokens, $0.001 per million cached input tokens and $0.14 per million output tokens, the same price at every hour. A free tier stays after that, on every account and with no card: at least 5,000 input and 1,000 output tokens per hour. We'll email every account before the switch, and nothing is charged until you add a payment method. How that compares with every other offer.
Current pricing
- deepseek-v4.1-flash-1017
- input / 1M tokens
- $0.005$0.00
- cached input / 1M
- $0.001$0.00
- output / 1M tokens
- $0.14$0.00
- rate limit
- none
- free tier after beta
- 5,000 in / 1,000 out per hour
- card needed
- no
What would you pay once the beta ends? Slide to your usage.
Frequently asked questions.
Is it really free?
Can I try it without writing code?
What does "no rate limit" mean, exactly?
503 with Retry-After, and your SDK's normal retry takes over.Will my existing OpenAI code work?
/v1/chat/completions, yes. Set base_url to our address, use your sk-cheap- key and set the model to deepseek-v4.1-flash-1017. temperature, top_p, max_tokens, stop, tools and response_format go straight to the model.Do you keep my prompts?
Can I see the model think?
delta.reasoning_content (mirrored as delta.reasoning) and the answer in delta.content.Which apps can I plug it into?
How do I keep keys under control?
What if I want out?
Get an API key in under a minute.
Email and password. Your key is shown right after signup, with a ready-to-run curl command.
Get a free API key Try the playground