openai 0.117.0 → 0.118.0

data/rbi/openai/models/responses/response_usage.rbi

@@ -108,13 +108,26 @@ module OpenAI
           sig { returns(Integer) }
           attr_accessor :cached_tokens
 
+          sig { returns(T.nilable(OpenAI::Responses::ResponseUsage::InputTokensDetails::CacheWriteTokensDetails)) }
+          attr_reader :cache_write_tokens_details
+
+          sig {
+            params(
+              cache_write_tokens_details: OpenAI::Responses::ResponseUsage::InputTokensDetails::CacheWriteTokensDetails::OrHash
+            )
+              .void
+          }
+          attr_writer :cache_write_tokens_details
+
           # A detailed breakdown of the input tokens.
           sig do
             params(
 
               cache_write_tokens: Integer,
 
-              cached_tokens: Integer
+              cached_tokens: Integer,
+
+              cache_write_tokens_details: OpenAI::Responses::ResponseUsage::InputTokensDetails::CacheWriteTokensDetails::OrHash
             )
               .returns(T.attached_class)
           end
@@ -125,19 +138,69 @@ module OpenAI
 
             # The number of tokens that were retrieved from the cache.
             # [More on prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching).
+            cached_tokens:,
 
-            cached_tokens:
+            cache_write_tokens_details: nil
           )
           end
 
           sig do
             override.returns(
-              {cache_write_tokens: Integer, cached_tokens: Integer}
+              {
+                cache_write_tokens: Integer,
+                cached_tokens: Integer,
+                cache_write_tokens_details: OpenAI::Responses::ResponseUsage::InputTokensDetails::CacheWriteTokensDetails
+              }
             )
           end
           def to_hash
           end
 
+          class CacheWriteTokensDetails < OpenAI::Internal::Type::BaseModel
+            OrHash = T.type_alias do
+              T.any(
+                OpenAI::Responses::ResponseUsage::InputTokensDetails::CacheWriteTokensDetails,
+                OpenAI::Internal::AnyHash
+              )
+            end
+
+            # The number of input tokens written with 12-hour retention.
+            sig { returns(Integer) }
+            attr_accessor :ttl_12h_tokens
+
+            # The number of input tokens written with 30-minute retention.
+            sig { returns(Integer) }
+            attr_accessor :ttl_30m_tokens
+
+            sig do
+              params(
+
+                ttl_12h_tokens: Integer,
+
+                ttl_30m_tokens: Integer
+              )
+                .returns(T.attached_class)
+            end
+            def self.new(
+
+              # The number of input tokens written with 12-hour retention.
+              ttl_12h_tokens:,
+
+              # The number of input tokens written with 30-minute retention.
+
+              ttl_30m_tokens:
+            )
+            end
+
+            sig do
+              override.returns(
+                {ttl_12h_tokens: Integer, ttl_30m_tokens: Integer}
+              )
+            end
+            def to_hash
+            end
+
+          end
         end
 
         class OutputTokensDetails < OpenAI::Internal::Type::BaseModel

data/rbi/openai/models/responses/responses_client_event.rbi

@@ -194,7 +194,7 @@ module OpenAI
           # request can write up to four breakpoints. For cache matching, OpenAI considers
           # up to the latest 80 breakpoints in the conversation, without a content-block
           # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-          # `ttl` defaults to `30m`, which is currently the only supported value. See the
+          # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           sig { returns(T.nilable(OpenAI::Responses::ResponsesClientEvent::ResponseCreate::PromptCacheOptions)) }
@@ -714,7 +714,7 @@ module OpenAI
             # request can write up to four breakpoints. For cache matching, OpenAI considers
             # up to the latest 80 breakpoints in the conversation, without a content-block
             # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-            # `ttl` defaults to `30m`, which is currently the only supported value. See the
+            # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
             # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
             # for current details.
             prompt_cache_options: nil,
@@ -1478,8 +1478,8 @@ module OpenAI
             attr_writer :prewarm
 
             # The minimum lifetime applied to every implicit and explicit cache breakpoint
-            # written by the request. Defaults to `30m`, which is currently the only supported
-            # value. The backend may retain cache entries for longer.
+            # written by the request. Supported values are `30m` (the default) and `12h`. The
+            # backend may retain cache entries for longer.
             sig {
               returns(
                 T.nilable(OpenAI::Responses::ResponsesClientEvent::ResponseCreate::PromptCacheOptions::Ttl::OrSymbol)
@@ -1499,7 +1499,7 @@ module OpenAI
             # request can write up to four breakpoints. For cache matching, OpenAI considers
             # up to the latest 80 breakpoints in the conversation, without a content-block
             # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-            # `ttl` defaults to `30m`, which is currently the only supported value. See the
+            # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
             # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
             # for current details.
             sig do
@@ -1534,8 +1534,8 @@ module OpenAI
               prewarm: nil,
 
               # The minimum lifetime applied to every implicit and explicit cache breakpoint
-              # written by the request. Defaults to `30m`, which is currently the only supported
-              # value. The backend may retain cache entries for longer.
+              # written by the request. Supported values are `30m` (the default) and `12h`. The
+              # backend may retain cache entries for longer.
 
               ttl: nil
             )
@@ -1593,8 +1593,8 @@ module OpenAI
             end
 
             # The minimum lifetime applied to every implicit and explicit cache breakpoint
-            # written by the request. Defaults to `30m`, which is currently the only supported
-            # value. The backend may retain cache entries for longer.
+            # written by the request. Supported values are `30m` (the default) and `12h`. The
+            # backend may retain cache entries for longer.
             module Ttl
               extend OpenAI::Internal::Type::Enum
 
@@ -1607,6 +1607,10 @@ module OpenAI
                 :"30m",
                 OpenAI::Responses::ResponsesClientEvent::ResponseCreate::PromptCacheOptions::Ttl::TaggedSymbol
               )
+              TTL_12H = T.let(
+                :"12h",
+                OpenAI::Responses::ResponsesClientEvent::ResponseCreate::PromptCacheOptions::Ttl::TaggedSymbol
+              )
 
               sig {
                 override.returns(

data/rbi/openai/resources/admin/organization/projects/service_tier_settings.rbi

@@ -0,0 +1,67 @@
+# typed: strong
+
+module OpenAI
+  module Resources
+
+    class Admin
+
+      class Organization
+
+        class Projects
+
+          class ServiceTierSettings
+
+            # Retrieve the configured default and enabled service tiers for a project.
+            sig {
+              params(project_id: String, request_options: OpenAI::RequestOptions::OrHash).returns(
+                OpenAI::Admin::Organization::Projects::ProjectServiceTierSettings
+              )
+            }
+            def retrieve(
+              # The ID of the project
+              project_id,
+              request_options: {}
+            )
+            end
+
+            # Update the default or enabled service tiers atomically. Supply at least one
+            # field. Omitted fields keep their current values.
+            sig {
+              params(
+                project_id: String,
+                default: OpenAI::Admin::Organization::Projects::ServiceTierSettingUpdateParams::Default::OrSymbol,
+                enabled_tiers: T.any(
+                  Symbol,
+                  T::Array[OpenAI::Admin::Organization::Projects::ProjectOptionalServiceTier::OrSymbol]
+                ),
+                request_options: OpenAI::RequestOptions::OrHash
+              )
+                .returns(OpenAI::Admin::Organization::Projects::ProjectServiceTierSettings)
+            }
+            def update(
+              # The ID of the project
+              project_id,
+              # Default tier for requests that omit service_tier or use auto.
+              default: nil,
+              # Additional service tiers enabled for this project. "all" includes future tiers.
+              # An empty array allows only Standard. Organization policy and eligibility still
+              # apply.
+              enabled_tiers: nil,
+              request_options: {}
+            )
+            end
+
+            # @api private
+            sig { params(client: OpenAI::Client).returns(T.attached_class) }
+            def self.new(client:)
+            end
+          end
+
+        end
+
+      end
+
+    end
+
+  end
+end

data/rbi/openai/resources/admin/organization/projects.rbi

@@ -45,6 +45,9 @@ module OpenAI
           sig { returns(OpenAI::Resources::Admin::Organization::Projects::Certificates) }
           attr_reader :certificates
 
+          sig { returns(OpenAI::Resources::Admin::Organization::Projects::ServiceTierSettings) }
+          attr_reader :service_tier_settings
+
           # Create a new project in the organization. Projects can be created and archived,
           # but cannot be deleted.
           sig {

data/rbi/openai/resources/beta/responses.rbi

@@ -218,8 +218,8 @@ module OpenAI
           # `prompt_cache_breakpoint`. Each request can write up to four breakpoints. For
           # cache matching, OpenAI considers up to the latest 80 breakpoints in the
           # conversation, without a content-block lookback limit. Set `mode` to `explicit`
-          # to disable the implicit breakpoint. The `ttl` defaults to `30m`, which is
-          # currently the only supported value. See the
+          # to disable the implicit breakpoint. The `ttl` defaults to `30m`. Supported
+          # lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           prompt_cache_options: nil,
@@ -540,8 +540,8 @@ module OpenAI
           # `prompt_cache_breakpoint`. Each request can write up to four breakpoints. For
           # cache matching, OpenAI considers up to the latest 80 breakpoints in the
           # conversation, without a content-block lookback limit. Set `mode` to `explicit`
-          # to disable the implicit breakpoint. The `ttl` defaults to `30m`, which is
-          # currently the only supported value. See the
+          # to disable the implicit breakpoint. The `ttl` defaults to `30m`. Supported
+          # lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           prompt_cache_options: nil,
@@ -852,8 +852,8 @@ module OpenAI
           # `prompt_cache_breakpoint`. Each request can write up to four breakpoints. For
           # cache matching, OpenAI considers up to the latest 80 breakpoints in the
           # conversation, without a content-block lookback limit. Set `mode` to `explicit`
-          # to disable the implicit breakpoint. The `ttl` defaults to `30m`, which is
-          # currently the only supported value. See the
+          # to disable the implicit breakpoint. The `ttl` defaults to `30m`. Supported
+          # lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           prompt_cache_options: nil,

data/rbi/openai/resources/chat/completions.rbi

@@ -222,7 +222,7 @@ module OpenAI
           # request can write up to four breakpoints. For cache matching, OpenAI considers
           # up to the latest 80 breakpoints in the conversation, without a content-block
           # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-          # `ttl` defaults to `30m`, which is currently the only supported value. See the
+          # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           prompt_cache_options: nil,
@@ -567,7 +567,7 @@ module OpenAI
           # request can write up to four breakpoints. For cache matching, OpenAI considers
           # up to the latest 80 breakpoints in the conversation, without a content-block
           # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-          # `ttl` defaults to `30m`, which is currently the only supported value. See the
+          # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
           # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
           # for current details.
           prompt_cache_options: nil,

data/rbi/openai/resources/responses.rbi

@@ -210,7 +210,7 @@ module OpenAI
         # request can write up to four breakpoints. For cache matching, OpenAI considers
         # up to the latest 80 breakpoints in the conversation, without a content-block
         # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-        # `ttl` defaults to `30m`, which is currently the only supported value. See the
+        # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
         # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
         # for current details.
         prompt_cache_options: nil,
@@ -521,7 +521,7 @@ module OpenAI
         # request can write up to four breakpoints. For cache matching, OpenAI considers
         # up to the latest 80 breakpoints in the conversation, without a content-block
         # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-        # `ttl` defaults to `30m`, which is currently the only supported value. See the
+        # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
         # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
         # for current details.
         prompt_cache_options: nil,
@@ -815,7 +815,7 @@ module OpenAI
         # request can write up to four breakpoints. For cache matching, OpenAI considers
         # up to the latest 80 breakpoints in the conversation, without a content-block
         # lookback limit. Set `mode` to `explicit` to disable the implicit breakpoint. The
-        # `ttl` defaults to `30m`, which is currently the only supported value. See the
+        # `ttl` defaults to `30m`. Supported lifetimes depend on the API. See the
         # [prompt caching guide](https://developers.openai.com/api/docs/guides/prompt-caching)
         # for current details.
         prompt_cache_options: nil,

data/sig/openai/models/admin/organization/projects/project_optional_service_tier.rbs

@@ -0,0 +1,23 @@
+module OpenAI
+  module Models
+    module Admin
+      module Organization
+        module ProjectOptionalServiceTier = Projects::ProjectOptionalServiceTier
+
+        module Projects
+          type project_optional_service_tier = :flex | :fast | :ultrafast
+
+          module ProjectOptionalServiceTier
+            extend OpenAI::Internal::Type::Enum
+
+            FLEX: :flex
+            FAST: :fast
+            ULTRAFAST: :ultrafast
+
+            def self?.values: -> ::Array[OpenAI::Models::Admin::Organization::Projects::project_optional_service_tier]
+          end
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/admin/organization/projects/project_service_tier_settings.rbs

@@ -0,0 +1,72 @@
+module OpenAI
+  module Models
+    module Admin
+      module Organization
+        class ProjectServiceTierSettings = Projects::ProjectServiceTierSettings
+
+        module Projects
+          type project_service_tier_settings =
+            {
+              default: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::default,
+              enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::enabled_tiers,
+              ?object: :"project.service_tier_settings"
+            }
+
+          class ProjectServiceTierSettings < OpenAI::Internal::Type::BaseModel
+            # Default tier for requests that omit service_tier or use auto.
+            attr_accessor default: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::default
+
+            # Additional service tiers enabled for this project. "all" includes future tiers.
+            # An empty array allows only Standard. Organization policy and eligibility still
+            # apply.
+            attr_accessor enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::enabled_tiers
+
+            attr_accessor object: :"project.service_tier_settings"
+
+            # @param default
+            #   Default tier for requests that omit service_tier or use auto.
+            # @param enabled_tiers
+            #   Additional service tiers enabled for this project. "all" includes future tiers.
+            #   An empty array allows only Standard. Organization policy and eligibility still
+            #   apply.
+            def initialize: (
+              default: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::default,
+              enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::enabled_tiers,
+              ?object: :"project.service_tier_settings"
+            ) -> void
+
+            def to_hash: -> {
+              default: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::default,
+              enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::enabled_tiers,
+              ?object: :"project.service_tier_settings"
+            }
+
+            type default = :standard | :fast | :scale
+
+            module Default
+              extend OpenAI::Internal::Type::Enum
+
+              STANDARD: :standard
+              FAST: :fast
+              SCALE: :scale
+
+              def self?.values: -> ::Array[OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::default]
+            end
+
+            type enabled_tiers =
+              :all
+              | ::Array[OpenAI::Models::Admin::Organization::Projects::project_optional_service_tier]
+
+            module EnabledTiers
+              extend OpenAI::Internal::Type::Union
+
+              def self?.variants: -> ::Array[OpenAI::Models::Admin::Organization::Projects::ProjectServiceTierSettings::enabled_tiers]
+
+              ProjectOptionalServiceTierArray: OpenAI::Internal::Type::Converter
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/admin/organization/projects/service_tier_setting_retrieve_params.rbs

@@ -0,0 +1,32 @@
+module OpenAI
+  module Models
+    module Admin
+      module Organization
+        module Projects
+          type service_tier_setting_retrieve_params =
+            { project_id: String, ?request_options: OpenAI::request_opts }
+
+          class ServiceTierSettingRetrieveParams < OpenAI::Internal::Type::BaseModel
+            extend OpenAI::Internal::Type::RequestParameters::Converter
+            include OpenAI::Internal::Type::RequestParameters
+
+            # The ID of the project
+            attr_accessor project_id: String
+
+            # @param project_id
+            #   The ID of the project
+            def initialize: (
+              project_id: String,
+              ?request_options: OpenAI::request_opts
+            ) -> void
+
+            def to_hash: -> {
+              project_id: String,
+              ?request_options: OpenAI::RequestOptions
+            }
+          end
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/admin/organization/projects/service_tier_setting_update_params.rbs

@@ -0,0 +1,91 @@
+module OpenAI
+  module Models
+    module Admin
+      module Organization
+        module Projects
+          type service_tier_setting_update_params =
+            {
+              project_id: String,
+              ?default: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default,
+              ?enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers,
+              ?request_options: OpenAI::request_opts
+            }
+
+          class ServiceTierSettingUpdateParams < OpenAI::Internal::Type::BaseModel
+            extend OpenAI::Internal::Type::RequestParameters::Converter
+            include OpenAI::Internal::Type::RequestParameters
+
+            # The ID of the project
+            attr_accessor project_id: String
+
+            # Default tier for requests that omit service_tier or use auto.
+            attr_reader default: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default?
+
+            # Default tier for requests that omit service_tier or use auto.
+            def default=: (
+              OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default
+            ) -> OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default
+
+            # Additional service tiers enabled for this project. "all" includes future tiers.
+            # An empty array allows only Standard. Organization policy and eligibility still
+            # apply.
+            attr_reader enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers?
+
+            # Additional service tiers enabled for this project. "all" includes future tiers.
+            # An empty array allows only Standard. Organization policy and eligibility still
+            # apply.
+            def enabled_tiers=: (
+              OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers
+            ) -> OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers
+
+            # @param project_id
+            #   The ID of the project
+            # @param default
+            #   Default tier for requests that omit service_tier or use auto.
+            # @param enabled_tiers
+            #   Additional service tiers enabled for this project. "all" includes future tiers.
+            #   An empty array allows only Standard. Organization policy and eligibility still
+            #   apply.
+            def initialize: (
+              project_id: String,
+              ?default: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default,
+              ?enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers,
+              ?request_options: OpenAI::request_opts
+            ) -> void
+
+            def to_hash: -> {
+              project_id: String,
+              ?default: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default,
+              ?enabled_tiers: OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers,
+              ?request_options: OpenAI::RequestOptions
+            }
+
+            type default = :standard | :fast | :scale
+
+            module Default
+              extend OpenAI::Internal::Type::Enum
+
+              STANDARD: :standard
+              FAST: :fast
+              SCALE: :scale
+
+              def self?.values: -> ::Array[OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::default]
+            end
+
+            type enabled_tiers =
+              :all
+              | ::Array[OpenAI::Models::Admin::Organization::Projects::project_optional_service_tier]
+
+            module EnabledTiers
+              extend OpenAI::Internal::Type::Union
+
+              def self?.variants: -> ::Array[OpenAI::Models::Admin::Organization::Projects::ServiceTierSettingUpdateParams::enabled_tiers]
+
+              ProjectOptionalServiceTierArray: OpenAI::Internal::Type::Converter
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/sig/openai/models/beta/beta_response.rbs

@@ -1414,12 +1414,13 @@ module OpenAI
             def self?.values: -> ::Array[OpenAI::Models::Beta::BetaResponse::PromptCacheOptions::mode]
           end
 
-          type ttl = :"30m"
+          type ttl = :"30m" | :"12h"
 
           module Ttl
             extend OpenAI::Internal::Type::Enum
 
             TTL_30M: :"30m"
+            TTL_12H: :"12h"
 
             def self?.values: -> ::Array[OpenAI::Models::Beta::BetaResponse::PromptCacheOptions::ttl]
           end

data/sig/openai/models/beta/beta_response_usage.rbs

@@ -55,7 +55,11 @@ module OpenAI
         }
 
         type input_tokens_details =
-          { cache_write_tokens: Integer, cached_tokens: Integer }
+          {
+            cache_write_tokens: Integer,
+            cached_tokens: Integer,
+            ?cache_write_tokens_details: OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails
+          }
 
         class InputTokensDetails < OpenAI::Internal::Type::BaseModel
           # The number of input tokens that were written to the cache.
@@ -65,6 +69,12 @@ module OpenAI
           # [More on prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching).
           attr_accessor cached_tokens: Integer
 
+          attr_reader cache_write_tokens_details: OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails?
+
+          def cache_write_tokens_details=: (
+            OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails
+          ) -> OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails
+
           # @param cache_write_tokens
           #   The number of input tokens that were written to the cache.
           # @param cached_tokens
@@ -72,13 +82,40 @@ module OpenAI
           #   [More on prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching).
           def initialize: (
             cache_write_tokens: Integer,
-            cached_tokens: Integer
+            cached_tokens: Integer,
+            ?cache_write_tokens_details: OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails
           ) -> void
 
           def to_hash: -> {
             cache_write_tokens: Integer,
-            cached_tokens: Integer
+            cached_tokens: Integer,
+            ?cache_write_tokens_details: OpenAI::Beta::BetaResponseUsage::InputTokensDetails::CacheWriteTokensDetails
           }
+
+          type cache_write_tokens_details =
+            { :ttl_12h_tokens => Integer, :ttl_30m_tokens => Integer }
+
+          class CacheWriteTokensDetails < OpenAI::Internal::Type::BaseModel
+            # The number of input tokens written with 12-hour retention.
+            attr_accessor ttl_12h_tokens: Integer
+
+            # The number of input tokens written with 30-minute retention.
+            attr_accessor ttl_30m_tokens: Integer
+
+            # @param ttl_12h_tokens
+            #   The number of input tokens written with 12-hour retention.
+            # @param ttl_30m_tokens
+            #   The number of input tokens written with 30-minute retention.
+            def initialize: (
+              ttl_12h_tokens: Integer,
+              ttl_30m_tokens: Integer
+            ) -> void
+
+            def to_hash: -> {
+              :ttl_12h_tokens => Integer,
+              :ttl_30m_tokens => Integer
+            }
+          end
         end
 
         type output_tokens_details = { reasoning_tokens: Integer }