Token usage always shows 0 when using Zhipu Coding Plan API with @ai-sdk/openai-compatible #25986

Description

When using @ai-sdk/openai-compatible with Zhipu's Coding Plan endpoint, the status bar always displays Context: 0% used and token cost is never tracked — despite the upstream API returning valid usage data in both streaming and non-streaming responses.

Note: This is a different issue from #20885 which covers @ai-sdk/openai (Responses API). This issue is specific to @ai-sdk/openai-compatible (Chat Completions API) with Zhipu's Coding Plan endpoint.

Configuration

"zhipu-coding-plan": {
  "npm": "@ai-sdk/openai-compatible",
  "options": {
    "apiKey": "{env:ZHIPU_API_KEY}",
    "baseURL": "https://open.bigmodel.cn/api/coding/paas/v4"
  }
}

Steps to Reproduce

1. Configure @ai-sdk/openai-compatible with Zhipu Coding Plan endpoint (/api/coding/paas/v4)
2. Start a conversation with any model
3. Observe the status bar — Context: 0% used stays at zero

Verification

I verified the upstream API returns valid usage data via direct curl:

Non-streaming response

{
  "usage": {
    "completion_tokens": 15,
    "prompt_tokens": 6,
    "total_tokens": 21,
    "prompt_tokens_details": { "cached_tokens": 5 },
    "completion_tokens_details": { "reasoning_tokens": 1 }
  }
}

Streaming response (last chunk before [DONE])

{
  "usage": {
    "prompt_tokens": 6,
    "completion_tokens": 212,
    "total_tokens": 218,
    "prompt_tokens_details": { "cached_tokens": 5 },
    "completion_tokens_details": { "reasoning_tokens": 167 }
  }
}

Key Observations

1. Same SDK, different baseURL, different result — I have another provider using the same @ai-sdk/openai-compatible SDK pointing to Zhipu's standard endpoint (https://open.bigmodel.cn/api/paas/v4), and context %used displays correctly there (~44-59%). The only difference is the baseURL path (/api/coding/paas/v4 vs /api/paas/v4).

2. Session I/O shows correctly — The right panel displays 44K in / 166 out correctly, but the top bar Context: 0% used remains at zero.

3. The Coding Plan endpoint may return a slightly different response structure — possibly missing or extra fields in the streaming SSE format that causes the SDK to fail to extract usage data.

Environment

• OpenCode: v1.14.37 (latest as of 2026-05-05)
• SDK: @ai-sdk/openai-compatible
• Provider: Zhipu (智谱) Coding Plan API
• Endpoint: https://open.bigmodel.cn/api/coding/paas/v4

Related

• Token usage always shows 0 when using @ai-sdk/openai with custom baseURL (Responses API) #20885 — Similar issue with @ai-sdk/openai (Responses API)