hud-evals · solvemproblr · Jul 15, 2026 · Jul 16, 2026
diff --git a/cookbooks/codex-coding/codex_agent.py b/cookbooks/codex-coding/codex_agent.py
@@ -37,6 +37,7 @@
     "gpt-5.1",
     "gpt-5.3-codex",
     "gpt-5.4",
+    "gpt-5.6",
 }
 
 PROMPT_TEMPLATE = """You are a skilled software developer. Complete the following task:

diff --git a/docs/changelog.mdx b/docs/changelog.mdx
@@ -3,6 +3,17 @@ title: "Changelog"
 description: "Product updates and release notes for the HUD SDK and platform, covering new features, CLI commands, integrations, and UI changes by release date."
 ---
 
+<Update label="July 14, 2026">
+## GPT-5.6 Support
+
+- **GPT-5.6 family** — `gpt-5.6` (Sol), `gpt-5.6-terra`, and `gpt-5.6-luna` are available through the
+  HUD gateway and `create_agent`. `OpenAIAgent` now defaults to `gpt-5.6`.
+- **Reasoning controls** — GPT-5.6 reasoning passes through `OpenAIConfig.reasoning` unchanged: effort
+  levels `none` through `xhigh` and `max`, and `mode` (`standard` or `pro`).
+- **OpenAI SDK floor** — the SDK now requires `openai>=2.45.0` (Responses API function-call outputs use
+  the current content-item types).
+</Update>
+
 <Update label="May 6, 2026">
 ## Models, Tasksets, Templates & Sharing
 

diff --git a/docs/v6/reference/agents.mdx b/docs/v6/reference/agents.mdx
@@ -56,7 +56,7 @@ agent = ClaudeAgent(ClaudeConfig(model="claude-sonnet-4-5", max_steps=30))
 | Agent | Config | Default model |
 |-------|--------|---------------|
 | `ClaudeAgent` | `ClaudeConfig` | `claude-sonnet-4-6` |
-| `OpenAIAgent` | `OpenAIConfig` | `gpt-5.5` |
+| `OpenAIAgent` | `OpenAIConfig` | `gpt-5.6` |
 | `GeminiAgent` | `GeminiConfig` | `gemini-3-pro-preview` |
 | `OpenAIChatAgent` | `OpenAIChatConfig` | `gpt-5.4-mini` |
 | `ClaudeSDKAgent` | `ClaudeSDKConfig` | `claude-sonnet-4-6` |

diff --git a/hud/agents/openai/tools/base.py b/hud/agents/openai/tools/base.py
@@ -12,7 +12,6 @@
     ResponseInputFileContentParam,
     ResponseInputImageContentParam,
     ResponseInputTextContentParam,
-    ResponseInputTextParam,
 )
 from openai.types.responses.response_input_param import FunctionCallOutput, ResponseInputItemParam
 
@@ -76,7 +75,7 @@ def format_openai_result(call: MCPToolCall, result: MCPToolResult) -> ResponseIn
                 logger.warning("Unknown content block type: %s", type(block))
 
     if not output_items:
-        output_items.append(ResponseInputTextParam(type="input_text", text=""))
+        output_items.append(ResponseInputTextContentParam(type="input_text", text=""))
 
     return cast(
         "ResponseInputItemParam",

diff --git a/hud/agents/tests/test_openai_agent.py b/hud/agents/tests/test_openai_agent.py
@@ -11,7 +11,9 @@
 from openai.types.responses import ResponseOutputText
 
 from hud.agents.openai.agent import OpenAIAgent, OpenAIRunState
+from hud.agents.openai.tools.base import format_openai_result
 from hud.agents.types import OpenAIConfig
+from hud.types import MCPToolCall, MCPToolResult
 
 
 class FakeResponses:
@@ -39,6 +41,17 @@ def test_format_message_shapes_user_text() -> None:
     assert msg["role"] == "user"
 
 
+def test_format_openai_result_empty_output_emits_one_text_item() -> None:
+    # The Responses API rejects a function_call_output with an empty output
+    # list, so a contentless tool result must still produce one input_text item.
+    call = MCPToolCall(id="call_1", name="noop")
+    formatted = cast("dict[str, Any]", format_openai_result(call, MCPToolResult(content=[])))
+
+    assert formatted["type"] == "function_call_output"
+    assert formatted["call_id"] == "call_1"
+    assert formatted["output"] == [{"type": "input_text", "text": ""}]
+
+
 def _api_response(
     id: str, output: list[Any], usage: Any = None, incomplete_details: Any = None
 ) -> Any:

diff --git a/hud/agents/types.py b/hud/agents/types.py
@@ -103,7 +103,7 @@ class OpenAIConfig(AgentConfig):
     """Configuration for OpenAIAgent."""
 
     model_name: str = "OpenAI"
-    model: str = Field(default="gpt-5.5", validation_alias=_model_alias)
+    model: str = Field(default="gpt-5.6", validation_alias=_model_alias)
     max_output_tokens: int | None = None
     temperature: float | None = None
     reasoning: Any = None  # openai Reasoning

diff --git a/hud/cli/eval.py b/hud/cli/eval.py
@@ -119,6 +119,7 @@ class AgentPreset:
 _AGENT_PRESETS: list[AgentPreset] = [
     AgentPreset("Claude Sonnet 4.6", AgentType.CLAUDE, "claude-sonnet-4-6"),
     AgentPreset("Claude Opus 4.8", AgentType.CLAUDE, "claude-opus-4-8"),
+    AgentPreset("GPT-5.6", AgentType.OPENAI, "gpt-5.6"),
     AgentPreset("GPT-5.5", AgentType.OPENAI, "gpt-5.5"),
     AgentPreset("Gemini 3.1 Pro (Preview)", AgentType.GEMINI, "gemini-3.1-pro-preview"),
     AgentPreset(
@@ -176,7 +177,7 @@ class AgentPreset:
 # use_computer_beta = true
 
 [openai]
-# model = "gpt-5.5"
+# model = "gpt-5.6"
 # temperature = 0.7
 # max_output_tokens = 4096
 

diff --git a/hud/eval/tests/test_rollout.py b/hud/eval/tests/test_rollout.py
@@ -249,7 +249,9 @@ async def start_child():
 
     try:
         with pytest.raises(RuntimeError, match="startup boom"):
-            async with SubprocessRuntime(env_file, ready_timeout=2.0)(Task(env="leaky", id="noop")):
+            async with SubprocessRuntime(env_file, ready_timeout=30.0)(
+                Task(env="leaky", id="noop")
+            ):
                 pass
         pid = int(pid_file.read_text())
         assert await _wait_for_pid_inactive(pid)

diff --git a/hud/tests/test_version.py b/hud/tests/test_version.py
@@ -1,8 +1,10 @@
 from __future__ import annotations
 
+from importlib.metadata import version
+
 
 def test_import():
     """Test that the package can be imported."""
     import hud
 
-    assert hud.__version__ == "0.6.9"
+    assert hud.__version__ == version("hud")
diff --git a/pyproject.toml b/pyproject.toml
@@ -18,7 +18,7 @@ dependencies = [
     "mcp>=1.24.0,<2.0",
     "fastmcp==3.0.2",
     # For all inference agents
-    "openai==2.44.0",
+    "openai>=2.45.0",
     "anthropic>=0.78.0",
     "google-genai",
     # CLI dependencies