Cursor
Set up a custom API endpoint and understand Cursor's limits.
Use the custom endpoint for chat. Tab completion continues to use Cursor’s built-in models; Agent features depend on your Cursor version.
Current Inferway catalog
No verification record exists for deepseek/deepseek-v4.1-uncensored in this client; the configuration below is a reference.
Before you start
Install the named client using its official instructions. Pi here means the pi-mono coding agent; Hermes means Nous Research's Hermes Agent.
Once your Inferway account is activated, create a key under Console → API Keys. Use a dedicated key for this client; a project's monthly budget caps what it can spend. Both live and test prefixes use the same inference and billing path; test is an environment label, not a free or simulated inference mode.
Use an Inferway key for Inferway's endpoint. An OpenRouter key belongs to OpenRouter and cannot authenticate to this address.
Keep the key in the client's credential store or your local environment. Do not paste it into a prompt, screenshot, or committed configuration. These examples use bash/zsh on macOS or Linux.
Configure Cursor
- Open Cursor Settings → Models. In the OpenAI API key field, enter your Inferway API key and save or enable it. Do not enter the text INFERWAY_API_KEY: this UI field expects the actual credential.
- Enable Override OpenAI Base URL and enter https://api.inferway.ai/v1. If your version does not show this option, consult Cursor's current instructions before proceeding.
- Add a custom model with the exact ID deepseek/deepseek-v4.1-uncensored. Enable it and explicitly select it in a new chat; do not select Auto or a Cursor-managed model for this check.
Base URL: https://api.inferway.ai/v1 Model ID: deepseek/deepseek-v4.1-uncensored API key: <your Inferway API key>
Confirm the connection
- Start a fresh conversation with the configured provider and model. Ask: "Reply with exactly: connected". Confirm that text arrives without an API error.
- Open Console → Requests to see the request status and token usage.
- For coding tasks, ask the client to read a non-sensitive file. Review any proposed edits before applying them.
Cursor-specific limits
Cursor's official BYOK documentation covers chat models and describes OpenAI support as standard, non-reasoning chat. Tab completion uses Cursor's own models. Agent, Composer, edits, background requests, and Responses-based requests require separate compatibility checks.
The Base URL override can affect built-in chat models too. Turn the override off when returning to Cursor-managed chat models. If validation fails but a direct Inferway request works, check Cursor's requested model and endpoint instead of repeatedly rotating keys.
Cursor sends BYOK requests through its backend for prompt construction. Inferway's retention policy does not describe how Cursor handles your code and prompts; review both services' policies.
Frequently asked questions
Does the custom Base URL apply to Tab code completion? No. Cursor's OpenAI Base URL override applies strictly to Chat and Composer. Tab code completion continues to use Cursor's built-in models and independent infrastructure.
How do I switch back to Cursor default models? Open Cursor Settings -> Models, and disable the 'Override OpenAI Base URL' toggle, or clear the custom Base URL field to restore default routing.
Are streaming and tool calling supported? Unverified. Cursor connects via OpenAI-compatible endpoints, but because Cursor cannot be tested headlessly without GUI interactions, streaming, Composer multi-file edits, and tool actions have no automated verification receipt in this repository and are marked unverified.
What should I check if model validation fails? Test your API key and endpoint directly with curl. If curl succeeds, verify that the model ID in Cursor matches the published ID exactly as shown on the /models page, and that /v1 is not duplicated.
If a request fails
Authentication: check that an API key is present, active, and allowed to use the model. Reopen the client from the terminal containing the variable. Never print the full key when debugging.
Wrong endpoint or model: check for a duplicate /v1, use the exact published model ID, and confirm the client is using Chat Completions. A request to /responses or /completions cannot be repaired by changing the key.
Limits or timeout: check wallet balance and key limits, respect Retry-After, and reduce the prompt or output budget. Keep streaming enabled for long outputs. Avoid repeated agent retries while diagnosing.
When contacting support, share the client version, model ID, timestamp, and request ID if available. Redact credentials and source code.
Available Text Models & Pricing
All models below can be called directly in your client by specifying the model ID. Rates are the prices published for your market; data retention differs by model, see the Data Retention column.
Loading prices
| Model Name | Model ID | Context Window | Data Retention | Input Price | Output Price | Typical Session Cost |
|---|---|---|---|---|---|---|
| DeepSeek V4.1 Flash 1017 | deepseek/deepseek-v4.1-flash-1017 | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
| DeepSeek V4.1 Flash 1017 ZDR | deepseek/deepseek-v4.1-flash-1017-zdr | 1,000,000 | Provider states zero data retention | — | — | — |
| DeepSeek V4.1 Flash Uncensored | deepseek/deepseek-v4.1-uncensored | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
| MiniMax M3.1 Flash | minimax/minimax-m3.1-flash | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
| Kimi K3.1 Pro | moonshotai/kimi-k3.1-pro | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
| MiMo-V2.6 Flash | xiaomi/mimo-v2.6-flash | 393,216 | Provider states zero data retention | — | — | — |
| GLM-5.3 | z-ai/glm-5.3 | 1,000,000 | Third-party provider; its retention terms apply | — | — | — |
| GLM-5.3 Flash Uncensored | z-ai/glm-5.3-flash-uncensored | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
| GLM-5.3 Uncensored | z-ai/glm-5.3-uncensored | 1,048,576 | Third-party provider; its retention terms apply | — | — | — |
Typical Coding Session Cost Example
Note: This example assumes a typical multi-turn coding session with 50,000 input tokens and 2,000 output tokens (input covers codebase context and turn history; output covers generated code edits and replies). Token counts are illustrative constants; unit prices are the rates published for your market (in that market's currency), and the total cost is calculated dynamically at render time. Actual charges depend on verified usage recorded in the console.