[feat] support o3-mini (#6570)

frontend/src/utils/verified-models.ts

@@ -1,7 +1,7 @@
 // Here are the list of verified models and providers that we know work well with OpenHands.
 export const VERIFIED_PROVIDERS = ["openai", "azure", "anthropic", "deepseek"];
 export const VERIFIED_MODELS = [
-  "gpt-4o",
+  "o3-mini-2025-01-31",
   "claude-3-5-sonnet-20241022",
   "deepseek-chat",
 ];
@@ -16,6 +16,8 @@ export const VERIFIED_OPENAI_MODELS = [
   "gpt-4-32k",
   "o1-mini",
   "o1-preview",
+  "o3-mini",
+  "o3-mini-2025-01-31",
 ];
 
 // LiteLLM does not return the compatible Anthropic models with the provider, so we list them here to set them ourselves

openhands/core/config/llm_config.py

@@ -85,7 +85,7 @@ class LLMConfig(BaseModel):
     log_completions_folder: str = Field(default=os.path.join(LOG_DIR, 'completions'))
     custom_tokenizer: str | None = Field(default=None)
     native_tool_calling: bool | None = Field(default=None)
-    reasoning_effort: str | None = Field(default=None)
+    reasoning_effort: str | None = Field(default='high')
 
     model_config = {'extra': 'forbid'}
 

openhands/llm/llm.py

@@ -71,6 +71,8 @@ FUNCTION_CALLING_SUPPORTED_MODELS = [
     'gpt-4o-mini',
     'gpt-4o',
     'o1-2024-12-17',
+    'o3-mini-2025-01-31',
+    'o3-mini',
 ]
 
 # visual browsing tool supported models
@@ -85,6 +87,9 @@ VISUAL_BROWSING_TOOL_SUPPORTED_MODELS = [
 
 REASONING_EFFORT_SUPPORTED_MODELS = [
     'o1-2024-12-17',
+    'o1',
+    'o3-mini-2025-01-31',
+    'o3-mini',
 ]
 
 MODELS_WITHOUT_STOP_WORDS = [
@@ -149,6 +154,18 @@ class LLM(RetryMixin, DebugMixin):
             self.tokenizer = None
 
         # set up the completion function
+        kwargs: dict[str, Any] = {
+            'temperature': self.config.temperature,
+        }
+        if (
+            self.config.model.lower() in REASONING_EFFORT_SUPPORTED_MODELS
+            or self.config.model.split('/')[-1] in REASONING_EFFORT_SUPPORTED_MODELS
+        ):
+            kwargs['reasoning_effort'] = self.config.reasoning_effort
+            kwargs.pop(
+                'temperature'
+            )  # temperature is not supported for reasoning models
+
         self._completion = partial(
             litellm_completion,
             model=self.config.model,
@@ -158,17 +175,11 @@ class LLM(RetryMixin, DebugMixin):
             base_url=self.config.base_url,
             api_version=self.config.api_version,
             custom_llm_provider=self.config.custom_llm_provider,
-            max_tokens=self.config.max_output_tokens,
+            max_completion_tokens=self.config.max_output_tokens,
             timeout=self.config.timeout,
-            temperature=self.config.temperature,
             top_p=self.config.top_p,
             drop_params=self.config.drop_params,
-            # add reasoning_effort, only if the model is supported
-            **(
-                {'reasoning_effort': self.config.reasoning_effort}
-                if self.config.model.lower() in REASONING_EFFORT_SUPPORTED_MODELS
-                else {}
-            ),
+            **kwargs,
         )
 
         self._completion_unwrapped = self._completion

openhands/runtime/utils/runtime_templates/Dockerfile.j2

@@ -61,7 +61,8 @@ RUN if [ -z "${RELEASE_TAG}" ]; then \
     fi && \
     wget https://github.com/${RELEASE_ORG}/openvscode-server/releases/download/${RELEASE_TAG}/${RELEASE_TAG}-linux-${arch}.tar.gz && \
     tar -xzf ${RELEASE_TAG}-linux-${arch}.tar.gz && \
-    mv -f ${RELEASE_TAG}-linux-${arch} ${OPENVSCODE_SERVER_ROOT} && \
+    if [ -d "${OPENVSCODE_SERVER_ROOT}" ]; then rm -rf "${OPENVSCODE_SERVER_ROOT}"; fi && \
+    mv ${RELEASE_TAG}-linux-${arch} ${OPENVSCODE_SERVER_ROOT} && \
     cp ${OPENVSCODE_SERVER_ROOT}/bin/remote-cli/openvscode-server ${OPENVSCODE_SERVER_ROOT}/bin/remote-cli/code && \
     rm -f ${RELEASE_TAG}-linux-${arch}.tar.gz && \
     # Install our custom extension