apiBase and model settings are being ignored

Before submitting your bug report

• I've tried finding an answer on the Continue docs site
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS: macOS
- Continue version: 2.0.0
- IDE version:  VS Code 1.126.0 (Universal)
- Model: DeepSeek
- config:
  
name: DeepSeek Config
version: 1.0.0
schema: v1
models:
  - name: DeepSeek V4 Flash
    provider: deepseek
    model: deepseek-v4-flash
    apiKey: ...
    apiBase: https://api.deepseek.com/v1
    roles:
      - chat
      - edit
      - apply
      - autocomplete

  - name: DeepSeek V4 Pro
    provider: deepseek
    model: deepseek-v4-pro
    apiKey: ...
    apiBase: https://api.deepseek.com/v1
    roles:
      - chat
      - edit
      - apply
    capabilities:
      - tool_use
  
  OR link to agent in Continue hub:

Description

Trying to use DeepSeek with Continue on VS Code.
curl test confirms my DeepSeek API key works. But when I prompt for model name and context window I get "I'm Claude 3.5 Sonnet (by Anthropic), and my context window is 200,000 tokens."

curl test
function ask() {
curl -s https://api.deepseek.com/v1/chat/completions
-H "Content-Type: application/json"
-H "Authorization: Bearer API_KEY"
-d "{"model":"deepseek-v4-flash","messages":[{"role":"user","content":"$*"}]}"
| jq -r '.choices[0].message.content'
}

ask "What model are you?"
你好！我是DeepSeek，由深度求索公司开发的AI助手。我是最新版本的DeepSeek模型，专门设计用来提供高质量的对话和问题解答服务。

我具备以下特点：

• 纯文本模型，但支持文件上传（图像、PDF、Word、Excel等）并从中读取文字信息
• 上下文长度1M，可以一次性处理像《三体》三部曲那样的大部头书籍
• 支持联网搜索（需要手动开启）
• 完全免费使用，无需付费

有什么我可以帮助你的吗？无论是回答问题、处理文件、还是进行创作，我都很乐意协助你！😊

To reproduce

No response

Log output

No source file or test is named in the report. Start by running the supplied curl for comparison, then trace how the v1 model configuration is loaded and how the selected provider request uses apiBase and model; done when the configured DeepSeek model is used in VS Code instead of the reported Claude response.