@@ -195,6 +195,7 @@ class "ResponsesApiParams" as src.models.common.responses.responses_api_params.R
   parallel_tool_calls : Optional[bool]
   previous_response_id : Optional[str]
   prompt : Optional[Prompt]
+  prompt_cache_key : Optional[str]
   reasoning : Optional[Reasoning]
   safety_identifier : Optional[str]
   store : bool
 
@@ -2231,13 +2231,14 @@
             },
             "ResponsesRequest": {
                 "additionalProperties": false,
-                "description": "Model representing a request for the Responses API following LCORE specification.\n\nAttributes:\n    input: Input text or structured input items containing the query.\n    model: Model identifier in format \"provider/model\". Auto-selected if not provided.\n    conversation: Conversation ID linking to an existing conversation. Accepts both\n        OpenAI and LCORE formats. Mutually exclusive with previous_response_id.\n    include: Explicitly specify output item types that are excluded by default but\n        should be included in the response.\n    instructions: System instructions or guidelines provided to the model (acts as\n        the system prompt).\n    max_infer_iters: Maximum number of inference iterations the model can perform.\n    max_output_tokens: Maximum number of tokens allowed in the response.\n    max_tool_calls: Maximum number of tool calls allowed in a single response.\n    metadata: Custom metadata dictionary with key-value pairs for tracking or logging.\n    parallel_tool_calls: Whether the model can make multiple tool calls in parallel.\n    previous_response_id: Identifier of the previous response in a multi-turn\n        conversation. Mutually exclusive with conversation.\n    prompt: Prompt object containing a template with variables for dynamic\n        substitution.\n    reasoning: Reasoning configuration for the response.\n    safety_identifier: Safety identifier for the response.\n    store: Whether to store the response in conversation history. Defaults to True.\n    stream: Whether to stream the response as it is generated. Defaults to False.\n    temperature: Sampling temperature controlling randomness (typically 0.0\u20132.0).\n    text: Text response configuration specifying output format constraints (JSON\n        schema, JSON object, or plain text).\n    tool_choice: Tool selection strategy (\"auto\", \"required\", \"none\", or specific\n        tool configuration).\n    tools: List of tools available to the model (file search, web search, function\n        calls, MCP tools). Defaults to all tools available to the model.\n    generate_topic_summary: LCORE-specific flag indicating whether to generate a\n        topic summary for new conversations. Defaults to True.\n    shield_ids: LCORE-specific list of configured shield names to apply.\n        If None, all configured shields are used.\n    solr: Optional Solr inline RAG options (mode, filters) or legacy filter-only dict.",
+                "description": "Model representing a request for the Responses API following LCORE specification.\n\nAttributes:\n    input: Input text or structured input items containing the query.\n    model: Model identifier in format \"provider/model\". Auto-selected if not provided.\n    conversation: Conversation ID linking to an existing conversation. Accepts both\n        OpenAI and LCORE formats. Mutually exclusive with previous_response_id.\n    include: Explicitly specify output item types that are excluded by default but\n        should be included in the response.\n    instructions: System instructions or guidelines provided to the model (acts as\n        the system prompt).\n    max_infer_iters: Maximum number of inference iterations the model can perform.\n    max_output_tokens: Maximum number of tokens allowed in the response.\n    max_tool_calls: Maximum number of tool calls allowed in a single response.\n    metadata: Custom metadata dictionary with key-value pairs for tracking or logging.\n    parallel_tool_calls: Whether the model can make multiple tool calls in parallel.\n    previous_response_id: Identifier of the previous response in a multi-turn\n        conversation. Mutually exclusive with conversation.\n    prompt_cache_key: Prompt cache key used by providers that support prompt\n        caching.\n    prompt: Prompt object containing a template with variables for dynamic\n        substitution.\n    reasoning: Reasoning configuration for the response.\n    safety_identifier: Safety identifier for the response.\n    store: Whether to store the response in conversation history. Defaults to True.\n    stream: Whether to stream the response as it is generated. Defaults to False.\n    temperature: Sampling temperature controlling randomness (typically 0.0\u20132.0).\n    text: Text response configuration specifying output format constraints (JSON\n        schema, JSON object, or plain text).\n    tool_choice: Tool selection strategy (\"auto\", \"required\", \"none\", or specific\n        tool configuration).\n    tools: List of tools available to the model (file search, web search, function\n        calls, MCP tools). Defaults to all tools available to the model.\n    generate_topic_summary: LCORE-specific flag indicating whether to generate a\n        topic summary for new conversations. Defaults to True.\n    shield_ids: LCORE-specific list of configured shield names to apply.\n        If None, all configured shields are used.\n    solr: Optional Solr inline RAG options (mode, filters) or legacy filter-only dict.",
                 "examples": [
                     {
                         "generate_topic_summary": true,
                         "input": "Hello World!",
                         "instructions": "You are a helpful assistant",
                         "model": "openai/gpt-5.6-luna",
+                        "prompt_cache_key": "example-cache-key",
                         "store": true,
                         "stream": false
                     }
@@ -2306,6 +2307,12 @@
                         "default": null,
                         "title": "Previous Response Id"
                     },
+                    "prompt_cache_key": {
+                        "type": "string",
+                        "nullable": true,
+                        "default": null,
+                        "title": "Prompt Cache Key"
+                    },
                     "prompt": {
                         "anyOf": [
                             {
 
@@ -954,6 +954,8 @@ Attributes:
     parallel_tool_calls: Whether the model can make multiple tool calls in parallel.
     previous_response_id: Identifier of the previous response in a multi-turn
         conversation. Mutually exclusive with conversation.
+    prompt_cache_key: Prompt cache key used by providers that support prompt
+        caching.
     prompt: Prompt object containing a template with variables for dynamic
         substitution.
     reasoning: Reasoning configuration for the response.
@@ -987,6 +989,7 @@ Attributes:
 | metadata | object |  |
 | parallel_tool_calls | boolean |  |
 | previous_response_id | string |  |
+| prompt_cache_key | string |  |
 | prompt |  |  |
 | reasoning |  |  |
 | safety_identifier | string |  |
 
@@ -82,6 +82,7 @@ class "ResponsesRequest" as src.models.api.requests.responses_openai.ResponsesRe
   parallel_tool_calls : Optional[bool]
   previous_response_id : Optional[str]
   prompt : Optional[Prompt]
+  prompt_cache_key : Optional[str]
   reasoning : Optional[Reasoning]
   safety_identifier : Optional[str]
   shield_ids : Optional[list[str]]