Qwen3-coder usage Error caused by non numeric data in the return value

[tianshangwuyun]

Initial Checks

• I confirm that I'm using the latest version of Pydantic AI
• I confirm that I searched for my issue in https://github.com/pydantic/pydantic-ai/issues before opening this issue

Description

my current version v0.4.4

I checked the latest version of the usage code and it also makes mistakes

Qwen3-coder "usage":{"completion_tokens":66,"prompt_tokens":22,"total_tokens":88,"completion_tokens_details":null,"prompt_tokens_details":{"audio_tokens":null,"cached_tokens":0,"cache_type":"implicit"}}

"cache_type":"implicit" Causing errors Usage.py incr function

this line

for key, value in incr_usage.details.items():
              self.details[key] = self.details.get(key, 0) + value

Example Code

Python, Pydantic AI & LLM client version

my current version v0.4.4 
I checked the latest version of the usage code and it also makes mistakes

model Qwen3-coder
LLM client openai

[Kludex]

Can you try the latest version of pydantic-ai? Also, can you share a code that I can run? Please provide all needed imports.

[DouweM]

@tianshangwuyun Please check the latest version of Pydantic AI.

Judging by the code, we're not reading all keys from prompt_tokens_details anymore as of #2378:

pydantic-ai/pydantic_ai_slim/pydantic_ai/models/openai.py

Lines 1376 to 1377 in 057c0bc

	u.input_audio_tokens = response_usage.prompt_tokens_details.audio_tokens or 0
	u.cache_read_tokens = response_usage.prompt_tokens_details.cached_tokens or 0