Migrate OpenAI models away from max_tokens to max_completion_tokens

pydantic_ai_slim/pydantic_ai/models/openai.py

@@ -279,7 +279,7 @@ async def _completions_create(
                 tool_choice=tool_choice or NOT_GIVEN,
                 stream=stream,
                 stream_options={'include_usage': True} if stream else NOT_GIVEN,
-                max_tokens=model_settings.get('max_tokens', NOT_GIVEN),
+                max_completion_tokens=model_settings.get('max_tokens', NOT_GIVEN),
                 temperature=model_settings.get('temperature', NOT_GIVEN),
                 top_p=model_settings.get('top_p', NOT_GIVEN),
                 timeout=model_settings.get('timeout', NOT_GIVEN),

tests/models/cassettes/test_openai/test_max_completion_tokens[gpt-4.5-preview].yaml

@@ -0,0 +1,78 @@
+interactions:
+- request:
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '123'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+    method: POST
+    parsed_body:
+      max_completion_tokens: 100
+      messages:
+      - content: hello
+        role: user
+      model: gpt-4.5-preview
+      n: 1
+      stream: false
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    headers:
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      connection:
+      - keep-alive
+      content-length:
+      - '807'
+      content-type:
+      - application/json
+      openai-organization:
+      - pydantic-28gund
+      openai-processing-ms:
+      - '1408'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      transfer-encoding:
+      - chunked
+    parsed_body:
+      choices:
+      - finish_reason: stop
+        index: 0
+        message:
+          annotations: []
+          content: Hello! How can I help you today?
+          refusal: null
+          role: assistant
+      created: 1742636225
+      id: chatcmpl-BDpZplWguNLn40wA5mIpCR3OIzvYP
+      model: gpt-4.5-preview-2025-02-27
+      object: chat.completion
+      service_tier: default
+      system_fingerprint: null
+      usage:
+        completion_tokens: 10
+        completion_tokens_details:
+          accepted_prediction_tokens: 0
+          audio_tokens: 0
+          reasoning_tokens: 0
+          rejected_prediction_tokens: 0
+        prompt_tokens: 8
+        prompt_tokens_details:
+          audio_tokens: 0
+          cached_tokens: 0
+        total_tokens: 18
+    status:
+      code: 200
+      message: OK
+version: 1

tests/models/cassettes/test_openai/test_max_completion_tokens[gpt-4o-mini].yaml

@@ -0,0 +1,79 @@
+interactions:
+- request:
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '119'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+    method: POST
+    parsed_body:
+      max_completion_tokens: 100
+      messages:
+      - content: hello
+        role: user
+      model: gpt-4o-mini
+      n: 1
+      stream: false
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    headers:
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      connection:
+      - keep-alive
+      content-length:
+      - '840'
+      content-type:
+      - application/json
+      openai-organization:
+      - pydantic-28gund
+      openai-processing-ms:
+      - '278'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      transfer-encoding:
+      - chunked
+    parsed_body:
+      choices:
+      - finish_reason: stop
+        index: 0
+        logprobs: null
+        message:
+          annotations: []
+          content: Hello! How can I assist you today?
+          refusal: null
+          role: assistant
+      created: 1742636224
+      id: chatcmpl-BDpZoxw4i90ZesN8iyrwLmGWRJ5lz
+      model: gpt-4o-mini-2024-07-18
+      object: chat.completion
+      service_tier: default
+      system_fingerprint: fp_b8bc95a0ac
+      usage:
+        completion_tokens: 10
+        completion_tokens_details:
+          accepted_prediction_tokens: 0
+          audio_tokens: 0
+          reasoning_tokens: 0
+          rejected_prediction_tokens: 0
+        prompt_tokens: 8
+        prompt_tokens_details:
+          audio_tokens: 0
+          cached_tokens: 0
+        total_tokens: 18
+    status:
+      code: 200
+      message: OK
+version: 1

tests/models/cassettes/test_openai/test_max_completion_tokens[o3-mini].yaml

@@ -0,0 +1,78 @@
+interactions:
+- request:
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '115'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+    method: POST
+    parsed_body:
+      max_completion_tokens: 100
+      messages:
+      - content: hello
+        role: user
+      model: o3-mini
+      n: 1
+      stream: false
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    headers:
+      access-control-expose-headers:
+      - X-Request-ID
+      alt-svc:
+      - h3=":443"; ma=86400
+      connection:
+      - keep-alive
+      content-length:
+      - '817'
+      content-type:
+      - application/json
+      openai-organization:
+      - pydantic-28gund
+      openai-processing-ms:
+      - '1895'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      transfer-encoding:
+      - chunked
+    parsed_body:
+      choices:
+      - finish_reason: stop
+        index: 0
+        message:
+          annotations: []
+          content: Hello there! How can I help you today?
+          refusal: null
+          role: assistant
+      created: 1742636222
+      id: chatcmpl-BDpZm1SiItIXIcDA0xRV9QGZhD97e
+      model: o3-mini-2025-01-31
+      object: chat.completion
+      service_tier: default
+      system_fingerprint: fp_617f206dd9
+      usage:
+        completion_tokens: 85
+        completion_tokens_details:
+          accepted_prediction_tokens: 0
+          audio_tokens: 0
+          reasoning_tokens: 64
+          rejected_prediction_tokens: 0
+        prompt_tokens: 7
+        prompt_tokens_details:
+          audio_tokens: 0
+          cached_tokens: 0
+        total_tokens: 92
+    status:
+      code: 200
+      message: OK
+version: 1

tests/models/test_openai.py

@@ -28,7 +28,7 @@
 from pydantic_ai.result import Usage
 from pydantic_ai.settings import ModelSettings
 
-from ..conftest import IsNow, TestEnv, raise_if_exception, try_import
+from ..conftest import IsNow, IsStr, TestEnv, raise_if_exception, try_import
 from .mock_async_stream import MockAsyncStream
 
 with try_import() as imports_successful:
@@ -684,3 +684,13 @@ def test_model_status_error(allow_model_requests: None) -> None:
     with pytest.raises(ModelHTTPError) as exc_info:
         agent.run_sync('hello')
     assert str(exc_info.value) == snapshot("status_code: 500, model_name: gpt-4o, body: {'error': 'test error'}")
+
+
+@pytest.mark.vcr()
+@pytest.mark.parametrize('model_name', ['o3-mini', 'gpt-4o-mini', 'gpt-4.5-preview'])
+async def test_max_completion_tokens(allow_model_requests: None, model_name: str, openai_api_key: str):
+    m = OpenAIModel(model_name, provider=OpenAIProvider(api_key=openai_api_key))
+    agent = Agent(m, model_settings=ModelSettings(max_tokens=100))
+
+    result = await agent.run('hello')
+    assert result.data == IsStr()