anthropic 0.77.1__py3-none-any.whl → 0.78.0__py3-none-any.whl

anthropic/resources/messages/messages.py

@@ -71,6 +71,8 @@ DEPRECATED_MODELS = {
     "claude-3-5-haiku-20241022": "February 19th, 2026",
 }
 
+MODELS_TO_WARN_WITH_THINKING_ENABLED = ["claude-opus-4-6"]
+
 
 class Messages(SyncAPIResource):
     @cached_property
@@ -103,6 +105,7 @@ class Messages(SyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -211,6 +214,9 @@ class Messages(SyncAPIResource):
               [models](https://docs.anthropic.com/en/docs/models-overview) for additional
               details and options.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -375,6 +381,7 @@ class Messages(SyncAPIResource):
         messages: Iterable[MessageParam],
         model: ModelParam,
         stream: Literal[True],
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -486,6 +493,9 @@ class Messages(SyncAPIResource):
 
               See [streaming](https://docs.claude.com/en/api/messages-streaming) for details.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -646,6 +656,7 @@ class Messages(SyncAPIResource):
         messages: Iterable[MessageParam],
         model: ModelParam,
         stream: bool,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -757,6 +768,9 @@ class Messages(SyncAPIResource):
 
               See [streaming](https://docs.claude.com/en/api/messages-streaming) for details.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -916,6 +930,7 @@ class Messages(SyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -947,6 +962,13 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         return self._post(
             "/v1/messages",
             body=maybe_transform(
@@ -954,6 +976,7 @@ class Messages(SyncAPIResource):
                     "max_tokens": max_tokens,
                     "messages": messages,
                     "model": model,
+                    "inference_geo": inference_geo,
                     "metadata": metadata,
                     "output_config": output_config,
                     "service_tier": service_tier,
@@ -985,6 +1008,7 @@ class Messages(SyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         output_format: None | JSONOutputFormatParam | type[ResponseFormatT] | Omit = omit,
@@ -1013,6 +1037,13 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper-Method": "stream",
             "X-Stainless-Stream-Helper": "messages",
@@ -1055,6 +1086,7 @@ class Messages(SyncAPIResource):
                     "max_tokens": max_tokens,
                     "messages": messages,
                     "model": model,
+                    "inference_geo": inference_geo,
                     "metadata": metadata,
                     "output_config": merged_output_config,
                     "container": container,
@@ -1121,6 +1153,13 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper": "messages.parse",
             **(extra_headers or {}),
@@ -1484,6 +1523,7 @@ class AsyncMessages(AsyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -1592,6 +1632,9 @@ class AsyncMessages(AsyncAPIResource):
               [models](https://docs.anthropic.com/en/docs/models-overview) for additional
               details and options.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -1756,6 +1799,7 @@ class AsyncMessages(AsyncAPIResource):
         messages: Iterable[MessageParam],
         model: ModelParam,
         stream: Literal[True],
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -1867,6 +1911,9 @@ class AsyncMessages(AsyncAPIResource):
 
               See [streaming](https://docs.claude.com/en/api/messages-streaming) for details.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -2027,6 +2074,7 @@ class AsyncMessages(AsyncAPIResource):
         messages: Iterable[MessageParam],
         model: ModelParam,
         stream: bool,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -2138,6 +2186,9 @@ class AsyncMessages(AsyncAPIResource):
 
               See [streaming](https://docs.claude.com/en/api/messages-streaming) for details.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           metadata: An object describing metadata about the request.
 
           output_config: Configuration options for the model's output, such as the output format.
@@ -2297,6 +2348,7 @@ class AsyncMessages(AsyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
@@ -2328,6 +2380,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         return await self._post(
             "/v1/messages",
             body=await async_maybe_transform(
@@ -2335,6 +2394,7 @@ class AsyncMessages(AsyncAPIResource):
                     "max_tokens": max_tokens,
                     "messages": messages,
                     "model": model,
+                    "inference_geo": inference_geo,
                     "metadata": metadata,
                     "output_config": output_config,
                     "service_tier": service_tier,
@@ -2366,6 +2426,7 @@ class AsyncMessages(AsyncAPIResource):
         max_tokens: int,
         messages: Iterable[MessageParam],
         model: ModelParam,
+        inference_geo: Optional[str] | Omit = omit,
         metadata: MetadataParam | Omit = omit,
         output_config: OutputConfigParam | Omit = omit,
         output_format: None | JSONOutputFormatParam | type[ResponseFormatT] | Omit = omit,
@@ -2394,6 +2455,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper-Method": "stream",
             "X-Stainless-Stream-Helper": "messages",
@@ -2435,6 +2503,7 @@ class AsyncMessages(AsyncAPIResource):
                     "max_tokens": max_tokens,
                     "messages": messages,
                     "model": model,
+                    "inference_geo": inference_geo,
                     "metadata": metadata,
                     "output_config": merged_output_config,
                     "container": container,
@@ -2501,6 +2570,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper": "messages.parse",
             **(extra_headers or {}),

anthropic/types/__init__.py

@@ -115,6 +115,7 @@ from .raw_content_block_start_event import RawContentBlockStartEvent as RawConte
 from .redacted_thinking_block_param import RedactedThinkingBlockParam as RedactedThinkingBlockParam
 from .thinking_config_enabled_param import ThinkingConfigEnabledParam as ThinkingConfigEnabledParam
 from .web_search_result_block_param import WebSearchResultBlockParam as WebSearchResultBlockParam
+from .thinking_config_adaptive_param import ThinkingConfigAdaptiveParam as ThinkingConfigAdaptiveParam
 from .thinking_config_disabled_param import ThinkingConfigDisabledParam as ThinkingConfigDisabledParam
 from .web_search_tool_20250305_param import WebSearchTool20250305Param as WebSearchTool20250305Param
 from .citation_content_block_location import CitationContentBlockLocation as CitationContentBlockLocation

anthropic/types/beta/__init__.py

@@ -34,8 +34,10 @@ from .beta_citation_config import BetaCitationConfig as BetaCitationConfig
 from .beta_citations_delta import BetaCitationsDelta as BetaCitationsDelta
 from .beta_signature_delta import BetaSignatureDelta as BetaSignatureDelta
 from .beta_web_fetch_block import BetaWebFetchBlock as BetaWebFetchBlock
+from .beta_compaction_block import BetaCompactionBlock as BetaCompactionBlock
 from .beta_container_params import BetaContainerParams as BetaContainerParams
 from .beta_input_json_delta import BetaInputJSONDelta as BetaInputJSONDelta
+from .beta_iterations_usage import BetaIterationsUsage as BetaIterationsUsage
 from .beta_text_block_param import BetaTextBlockParam as BetaTextBlockParam
 from .beta_tool_union_param import BetaToolUnionParam as BetaToolUnionParam
 from .message_create_params import MessageCreateParams as MessageCreateParams
@@ -72,6 +74,7 @@ from .beta_base64_pdf_block_param import BetaBase64PDFBlockParam as BetaBase64PD
 from .beta_citation_char_location import BetaCitationCharLocation as BetaCitationCharLocation
 from .beta_citation_page_location import BetaCitationPageLocation as BetaCitationPageLocation
 from .beta_citations_config_param import BetaCitationsConfigParam as BetaCitationsConfigParam
+from .beta_compaction_block_param import BetaCompactionBlockParam as BetaCompactionBlockParam
 from .beta_container_upload_block import BetaContainerUploadBlock as BetaContainerUploadBlock
 from .beta_raw_message_stop_event import BetaRawMessageStopEvent as BetaRawMessageStopEvent
 from .beta_tool_choice_auto_param import BetaToolChoiceAutoParam as BetaToolChoiceAutoParam
@@ -81,6 +84,7 @@ from .beta_url_image_source_param import BetaURLImageSourceParam as BetaURLImage
 from .message_count_tokens_params import MessageCountTokensParams as MessageCountTokensParams
 from .beta_base64_pdf_source_param import BetaBase64PDFSourceParam as BetaBase64PDFSourceParam
 from .beta_file_image_source_param import BetaFileImageSourceParam as BetaFileImageSourceParam
+from .beta_message_iteration_usage import BetaMessageIterationUsage as BetaMessageIterationUsage
 from .beta_plain_text_source_param import BetaPlainTextSourceParam as BetaPlainTextSourceParam
 from .beta_raw_content_block_delta import BetaRawContentBlockDelta as BetaRawContentBlockDelta
 from .beta_raw_message_delta_event import BetaRawMessageDeltaEvent as BetaRawMessageDeltaEvent
@@ -98,6 +102,7 @@ from .beta_tool_bash_20241022_param import BetaToolBash20241022Param as BetaTool
 from .beta_tool_bash_20250124_param import BetaToolBash20250124Param as BetaToolBash20250124Param
 from .beta_base64_image_source_param import BetaBase64ImageSourceParam as BetaBase64ImageSourceParam
 from .beta_search_result_block_param import BetaSearchResultBlockParam as BetaSearchResultBlockParam
+from .beta_compaction_iteration_usage import BetaCompactionIterationUsage as BetaCompactionIterationUsage
 from .beta_content_block_source_param import BetaContentBlockSourceParam as BetaContentBlockSourceParam
 from .beta_file_document_source_param import BetaFileDocumentSourceParam as BetaFileDocumentSourceParam
 from .beta_input_tokens_trigger_param import BetaInputTokensTriggerParam as BetaInputTokensTriggerParam
@@ -105,6 +110,7 @@ from .beta_memory_tool_20250818_param import BetaMemoryTool20250818Param as Beta
 from .beta_tool_reference_block_param import BetaToolReferenceBlockParam as BetaToolReferenceBlockParam
 from .beta_code_execution_output_block import BetaCodeExecutionOutputBlock as BetaCodeExecutionOutputBlock
 from .beta_code_execution_result_block import BetaCodeExecutionResultBlock as BetaCodeExecutionResultBlock
+from .beta_compact_20260112_edit_param import BetaCompact20260112EditParam as BetaCompact20260112EditParam
 from .beta_context_management_response import BetaContextManagementResponse as BetaContextManagementResponse
 from .beta_server_tool_use_block_param import BetaServerToolUseBlockParam as BetaServerToolUseBlockParam
 from .beta_web_fetch_tool_result_block import BetaWebFetchToolResultBlock as BetaWebFetchToolResultBlock
@@ -126,6 +132,8 @@ from .beta_tool_search_tool_result_block import BetaToolSearchToolResultBlock as
 from .beta_tool_search_tool_result_error import BetaToolSearchToolResultError as BetaToolSearchToolResultError
 from .beta_web_fetch_tool_20250910_param import BetaWebFetchTool20250910Param as BetaWebFetchTool20250910Param
 from .beta_web_search_result_block_param import BetaWebSearchResultBlockParam as BetaWebSearchResultBlockParam
+from .beta_compaction_content_block_delta import BetaCompactionContentBlockDelta as BetaCompactionContentBlockDelta
+from .beta_thinking_config_adaptive_param import BetaThinkingConfigAdaptiveParam as BetaThinkingConfigAdaptiveParam
 from .beta_thinking_config_disabled_param import BetaThinkingConfigDisabledParam as BetaThinkingConfigDisabledParam
 from .beta_web_search_tool_20250305_param import BetaWebSearchTool20250305Param as BetaWebSearchTool20250305Param
 from .beta_citation_content_block_location import BetaCitationContentBlockLocation as BetaCitationContentBlockLocation

anthropic/types/beta/beta_compact_20260112_edit_param.py

@@ -0,0 +1,27 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing import Optional
+from typing_extensions import Literal, Required, TypedDict
+
+from .beta_input_tokens_trigger_param import BetaInputTokensTriggerParam
+
+__all__ = ["BetaCompact20260112EditParam"]
+
+
+class BetaCompact20260112EditParam(TypedDict, total=False):
+    """
+    Automatically compact older context when reaching the configured trigger threshold.
+    """
+
+    type: Required[Literal["compact_20260112"]]
+
+    instructions: Optional[str]
+    """Additional instructions for summarization."""
+
+    pause_after_compaction: bool
+    """Whether to pause after compaction and return the compaction block to the user."""
+
+    trigger: Optional[BetaInputTokensTriggerParam]
+    """When to trigger compaction. Defaults to 150000 input tokens."""

anthropic/types/beta/beta_compaction_block.py

@@ -0,0 +1,22 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+
+__all__ = ["BetaCompactionBlock"]
+
+
+class BetaCompactionBlock(BaseModel):
+    """A compaction block returned when autocompact is triggered.
+
+    When content is None, it indicates the compaction failed to produce a valid
+    summary (e.g., malformed output from the model). Clients may round-trip
+    compaction blocks with null content; the server treats them as no-ops.
+    """
+
+    content: Optional[str] = None
+    """Summary of compacted content, or null if compaction failed"""
+
+    type: Literal["compaction"]

anthropic/types/beta/beta_compaction_block_param.py

@@ -0,0 +1,29 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing import Optional
+from typing_extensions import Literal, Required, TypedDict
+
+from .beta_cache_control_ephemeral_param import BetaCacheControlEphemeralParam
+
+__all__ = ["BetaCompactionBlockParam"]
+
+
+class BetaCompactionBlockParam(TypedDict, total=False):
+    """A compaction block containing summary of previous context.
+
+    Users should round-trip these blocks from responses to subsequent requests
+    to maintain context across compaction boundaries.
+
+    When content is None, the block represents a failed compaction. The server
+    treats these as no-ops. Empty string content is not allowed.
+    """
+
+    content: Required[Optional[str]]
+    """Summary of previously compacted content, or null if compaction failed"""
+
+    type: Required[Literal["compaction"]]
+
+    cache_control: Optional[BetaCacheControlEphemeralParam]
+    """Create a cache control breakpoint at this content block."""

anthropic/types/beta/beta_compaction_content_block_delta.py

@@ -0,0 +1,14 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+
+__all__ = ["BetaCompactionContentBlockDelta"]
+
+
+class BetaCompactionContentBlockDelta(BaseModel):
+    content: Optional[str] = None
+
+    type: Literal["compaction_delta"]

anthropic/types/beta/beta_compaction_iteration_usage.py

@@ -0,0 +1,31 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+from .beta_cache_creation import BetaCacheCreation
+
+__all__ = ["BetaCompactionIterationUsage"]
+
+
+class BetaCompactionIterationUsage(BaseModel):
+    """Token usage for a compaction iteration."""
+
+    cache_creation: Optional[BetaCacheCreation] = None
+    """Breakdown of cached tokens by TTL"""
+
+    cache_creation_input_tokens: int
+    """The number of input tokens used to create the cache entry."""
+
+    cache_read_input_tokens: int
+    """The number of input tokens read from the cache."""
+
+    input_tokens: int
+    """The number of input tokens which were used."""
+
+    output_tokens: int
+    """The number of output tokens which were used."""
+
+    type: Literal["compaction"]
+    """Usage for a compaction iteration"""

anthropic/types/beta/beta_content_block.py

@@ -7,6 +7,7 @@ from ..._utils import PropertyInfo
 from .beta_text_block import BetaTextBlock
 from .beta_thinking_block import BetaThinkingBlock
 from .beta_tool_use_block import BetaToolUseBlock
+from .beta_compaction_block import BetaCompactionBlock
 from .beta_mcp_tool_use_block import BetaMCPToolUseBlock
 from .beta_mcp_tool_result_block import BetaMCPToolResultBlock
 from .beta_server_tool_use_block import BetaServerToolUseBlock
@@ -37,6 +38,7 @@ BetaContentBlock: TypeAlias = Annotated[
         BetaMCPToolUseBlock,
         BetaMCPToolResultBlock,
         BetaContainerUploadBlock,
+        BetaCompactionBlock,
     ],
     PropertyInfo(discriminator="type"),
 ]

anthropic/types/beta/beta_content_block_param.py

@@ -10,6 +10,7 @@ from .beta_text_block_param import BetaTextBlockParam
 from .beta_image_block_param import BetaImageBlockParam
 from .beta_thinking_block_param import BetaThinkingBlockParam
 from .beta_tool_use_block_param import BetaToolUseBlockParam
+from .beta_compaction_block_param import BetaCompactionBlockParam
 from .beta_tool_result_block_param import BetaToolResultBlockParam
 from .beta_mcp_tool_use_block_param import BetaMCPToolUseBlockParam
 from .beta_search_result_block_param import BetaSearchResultBlockParam
@@ -46,5 +47,6 @@ BetaContentBlockParam: TypeAlias = Union[
     BetaMCPToolUseBlockParam,
     BetaRequestMCPToolResultBlockParam,
     BetaContainerUploadBlockParam,
+    BetaCompactionBlockParam,
     BetaContentBlock,
 ]

anthropic/types/beta/beta_context_management_config_param.py

@@ -5,12 +5,15 @@ from __future__ import annotations
 from typing import Union, Iterable
 from typing_extensions import TypeAlias, TypedDict
 
+from .beta_compact_20260112_edit_param import BetaCompact20260112EditParam
 from .beta_clear_thinking_20251015_edit_param import BetaClearThinking20251015EditParam
 from .beta_clear_tool_uses_20250919_edit_param import BetaClearToolUses20250919EditParam
 
 __all__ = ["BetaContextManagementConfigParam", "Edit"]
 
-Edit: TypeAlias = Union[BetaClearToolUses20250919EditParam, BetaClearThinking20251015EditParam]
+Edit: TypeAlias = Union[
+    BetaClearToolUses20250919EditParam, BetaClearThinking20251015EditParam, BetaCompact20260112EditParam
+]
 
 
 class BetaContextManagementConfigParam(TypedDict, total=False):

anthropic/types/beta/beta_iterations_usage.py

@@ -0,0 +1,13 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import List, Union
+from typing_extensions import TypeAlias
+
+from .beta_message_iteration_usage import BetaMessageIterationUsage
+from .beta_compaction_iteration_usage import BetaCompactionIterationUsage
+
+__all__ = ["BetaIterationsUsage", "BetaIterationsUsageItem"]
+
+BetaIterationsUsageItem: TypeAlias = Union[BetaMessageIterationUsage, BetaCompactionIterationUsage]
+
+BetaIterationsUsage: TypeAlias = List[BetaIterationsUsageItem]

anthropic/types/beta/beta_message_delta_usage.py

@@ -3,6 +3,7 @@
 from typing import Optional
 
 from ..._models import BaseModel
+from .beta_iterations_usage import BetaIterationsUsage
 from .beta_server_tool_usage import BetaServerToolUsage
 
 __all__ = ["BetaMessageDeltaUsage"]
@@ -18,6 +19,17 @@ class BetaMessageDeltaUsage(BaseModel):
     input_tokens: Optional[int] = None
     """The cumulative number of input tokens which were used."""
 
+    iterations: Optional[BetaIterationsUsage] = None
+    """Per-iteration token usage breakdown.
+
+    Each entry represents one sampling iteration, with its own input/output token
+    counts and cache statistics. This allows you to:
+
+    - Determine which iterations exceeded long context thresholds (>=200k tokens)
+    - Calculate the true context window size from the last iteration
+    - Understand token accumulation across server-side tool use loops
+    """
+
     output_tokens: int
     """The cumulative number of output tokens which were used."""
 

anthropic/types/beta/beta_message_iteration_usage.py

@@ -0,0 +1,31 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from typing import Optional
+from typing_extensions import Literal
+
+from ..._models import BaseModel
+from .beta_cache_creation import BetaCacheCreation
+
+__all__ = ["BetaMessageIterationUsage"]
+
+
+class BetaMessageIterationUsage(BaseModel):
+    """Token usage for a sampling iteration."""
+
+    cache_creation: Optional[BetaCacheCreation] = None
+    """Breakdown of cached tokens by TTL"""
+
+    cache_creation_input_tokens: int
+    """The number of input tokens used to create the cache entry."""
+
+    cache_read_input_tokens: int
+    """The number of input tokens read from the cache."""
+
+    input_tokens: int
+    """The number of input tokens which were used."""
+
+    output_tokens: int
+    """The number of output tokens which were used."""
+
+    type: Literal["message"]
+    """Usage for a sampling iteration"""

anthropic/types/beta/beta_output_config_param.py

@@ -11,13 +11,8 @@ __all__ = ["BetaOutputConfigParam"]
 
 
 class BetaOutputConfigParam(TypedDict, total=False):
-    effort: Optional[Literal["low", "medium", "high"]]
-    """How much effort the model should put into its response.
-
-    Higher effort levels may result in more thorough analysis but take longer.
-
-    Valid values are `low`, `medium`, or `high`.
-    """
+    effort: Optional[Literal["low", "medium", "high", "max"]]
+    """All possible effort levels."""
 
     format: Optional[BetaJSONOutputFormatParam]
     """A schema to specify Claude's output format in responses.

anthropic/types/beta/beta_raw_content_block_delta.py

@@ -9,10 +9,18 @@ from .beta_thinking_delta import BetaThinkingDelta
 from .beta_citations_delta import BetaCitationsDelta
 from .beta_signature_delta import BetaSignatureDelta
 from .beta_input_json_delta import BetaInputJSONDelta
+from .beta_compaction_content_block_delta import BetaCompactionContentBlockDelta
 
 __all__ = ["BetaRawContentBlockDelta"]
 
 BetaRawContentBlockDelta: TypeAlias = Annotated[
-    Union[BetaTextDelta, BetaInputJSONDelta, BetaCitationsDelta, BetaThinkingDelta, BetaSignatureDelta],
+    Union[
+        BetaTextDelta,
+        BetaInputJSONDelta,
+        BetaCitationsDelta,
+        BetaThinkingDelta,
+        BetaSignatureDelta,
+        BetaCompactionContentBlockDelta,
+    ],
     PropertyInfo(discriminator="type"),
 ]

anthropic/types/beta/beta_raw_content_block_start_event.py

@@ -8,6 +8,7 @@ from ..._models import BaseModel
 from .beta_text_block import BetaTextBlock
 from .beta_thinking_block import BetaThinkingBlock
 from .beta_tool_use_block import BetaToolUseBlock
+from .beta_compaction_block import BetaCompactionBlock
 from .beta_mcp_tool_use_block import BetaMCPToolUseBlock
 from .beta_mcp_tool_result_block import BetaMCPToolResultBlock
 from .beta_server_tool_use_block import BetaServerToolUseBlock
@@ -38,6 +39,7 @@ ContentBlock: TypeAlias = Annotated[
         BetaMCPToolUseBlock,
         BetaMCPToolResultBlock,
         BetaContainerUploadBlock,
+        BetaCompactionBlock,
     ],
     PropertyInfo(discriminator="type"),
 ]

anthropic/types/beta/beta_stop_reason.py

@@ -5,5 +5,12 @@ from typing_extensions import Literal, TypeAlias
 __all__ = ["BetaStopReason"]
 
 BetaStopReason: TypeAlias = Literal[
-    "end_turn", "max_tokens", "stop_sequence", "tool_use", "pause_turn", "refusal", "model_context_window_exceeded"
+    "end_turn",
+    "max_tokens",
+    "stop_sequence",
+    "tool_use",
+    "pause_turn",
+    "compaction",
+    "refusal",
+    "model_context_window_exceeded",
 ]

anthropic/types/beta/beta_thinking_config_adaptive_param.py

@@ -0,0 +1,11 @@
+# File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+from __future__ import annotations
+
+from typing_extensions import Literal, Required, TypedDict
+
+__all__ = ["BetaThinkingConfigAdaptiveParam"]
+
+
+class BetaThinkingConfigAdaptiveParam(TypedDict, total=False):
+    type: Required[Literal["adaptive"]]