Continue.dev with any AI model (₹ pricing)

Continue.dev + unoblox: code completion in your IDE, billed in rupees

Continue is an open-source copilot plugin for VS Code and JetBrains. It supports both OpenAI-compatible and Anthropic-compatible endpoints, so pointing it at unoblox is a config change, not a rewrite. Pick an open model like Qwen or DeepSeek for day-to-day autocomplete, keep Claude available for harder chat-panel questions, and get one monthly ₹ invoice with GST input credit — no foreign card, no separate accounts to manage.

Setup: edit .continue/config.json

{
  "models": [
    {
      "name": "qwen-fast",
      "provider": "openai",
      "model": "qwen/qwen3-235b-a22b",
      "apiBase": "https://api.unoblox.ai/v1",
      "apiKey": "ub-gw-YOUR_API_KEY"
    },
    {
      "name": "deepseek-reasoning",
      "provider": "openai",
      "model": "deepseek-ai/deepseek-v3.2",
      "apiBase": "https://api.unoblox.ai/v1",
      "apiKey": "ub-gw-YOUR_API_KEY"
    }
  ],
  "tabAutocompleteModel": "qwen-fast"
}

Restart VS Code or your JetBrains IDE; autocomplete and chat now run against the unoblox gateway.

Pricing: ₹ per token consumed

Model	Input ₹/1M	Output ₹/1M	Best for
Qwen3.8-27B	₹16.32	₹48.96	Everyday tab autocomplete
DeepSeek V3.2	₹26.21	₹38.30	Deeper chat-panel reasoning
Qwen3 1.7B	₹0	₹0	Free tier, rate-limited
Claude Sonnet	₹201.6	₹1008	Hardest refactors, highest quality

Choosing a model per task

Autocomplete fires constantly—every keystroke pause can trigger a completion request—so it should run on a cheap, fast model. Chat-panel questions ("why is this test failing", "refactor this class") happen far less often and benefit from a stronger model. A sensible default: wire qwen-fast (or the free Qwen3 1.7B tier) to tabAutocompleteModel, and select DeepSeek or Claude Sonnet manually in the chat panel when you need deeper reasoning.

IDE + model combinations

• VS Code: Any model works; both OpenAI-compatible and Anthropic-compatible /v1/* endpoints are supported.
• JetBrains (IntelliJ, PyCharm, WebStorm, etc.): Same config shape—autocomplete plus chat-panel refactoring.
• Tab autocompletion: Point it at a fast, cheap model (Qwen) so it never feels like it's in your way.
• Chat panel: Point it at a reasoning model (DeepSeek or Claude Sonnet) for questions that need more than a one-line completion.

Frequently asked questions

Q: Will autocomplete slow down my IDE? Autocomplete requests are small and run against a fast, lightweight model by default, so it should feel comparable to any cloud-based completion tool. Switch models on the free tier if you want to test cost before committing.

Q: Can I use multiple models in Continue? Yes. Define multiple model objects in config.json; select whichever one you need from the dropdown, or wire different models to autocomplete vs. chat.

Q: Does Continue send my code to unoblox? Yes, to the unoblox gateway only, to generate the completion or chat response. unoblox does not train models on your data.

Q: Can I run Continue locally with no internet? Not against unoblox—Continue needs to reach the gateway for any unoblox-hosted model. For fully offline use, see Continue's own docs for self-hosted local models.

Q: Will my setup persist across IDE upgrades? Yes. .continue/config.json is a plain file you can version-control alongside your repo, so it survives upgrades and syncs across your team.

Q: Can I use Continue with JetBrains refactoring (rename, extract method)? Yes. The chat panel works the same way in JetBrains as in VS Code, and tab autocompletion is supported across most JetBrains IDEs.

Get started in rupees → https://unoblox.ai/sign-in