Using GLM-5.2 and DeepSeek in Continue.dev (config.yaml guide)

Continue.dev is the open-source IDE assistant for VS Code and JetBrains, and its openai provider accepts a custom apiBase. So adding GLM-5.2 or DeepSeek is a few lines of config.yaml — no plugin, no code changes.

The config

Add a model block to your Continue config.yaml:

models:
  - name: GLM-5.2 (Turiloop)
    provider: openai
    apiBase: https://api.turiloop.com/v1
    apiKey: YOUR_TURILOOP_KEY
    model: glm-5.2
    roles:
      - chat
      - edit

Swap model for deepseek-v4-pro or deepseek-v4-flash to add more entries. The apiKey is sent as a Bearer token, exactly like OpenAI.

Picking roles per model

Continue splits work into roles — chat, edit, autocomplete — and you can point each at the model that fits:

• chat / edit → glm-5.2: strong reasoning and a 1M context for whole-file edits (deep dive).
• autocomplete → deepseek-v4-flash: autocomplete fires constantly, so the cheapest fast model wins. Flash is priced at a fraction of a cent per thousand tokens.

Give each role its own model block with the right roles: list and Continue routes automatically.

Why Turiloop

Turiloop is an OpenAI-compatible API gateway: one key for GLM-5.2, DeepSeek, Kimi and MiniMax, pay-as-you-go with an international card, no Chinese phone number. Because it speaks the OpenAI format, Continue's built-in openai provider works with no extra adapter. More in one key, every model.

FAQ

How do I add DeepSeek to Continue.dev? Add a model block with provider: openai, apiBase: https://api.turiloop.com/v1, and model: deepseek-v4-pro (or deepseek-v4-flash).

Can Continue use different models for chat and autocomplete? Yes — that is what the roles field is for. Put GLM-5.2 on chat/edit and DeepSeek V4-Flash on autocomplete.

Does GLM-5.2 support tool use in Continue? Yes; add tool_use under capabilities if your workflow needs it.

Create a key on Turiloop, drop it into config.yaml, and reload Continue.