AnthropicModelCapabilities
Namespace: Vercel.AI.Anthropic
public sealed class AnthropicModelCapabilitiesWhat a Claude model id accepts on the Messages API.
Inheritance: object
Constructors
Section titled “Constructors”AnthropicModelCapabilities(int, bool, bool, bool, bool, bool, bool, bool, bool, bool, bool)
Section titled “AnthropicModelCapabilities(int, bool, bool, bool, bool, bool, bool, bool, bool, bool, bool)”public AnthropicModelCapabilities(int maxOutputTokens, bool supportsStructuredOutput, bool supportsAdaptiveThinking, bool rejectsSamplingParameters, bool supportsXhighEffort, bool rejectsThinkingDisabledAboveHighEffort, bool rejectsThinkingDisabled, bool rejectsBudgetThinking, bool rejectsForcedToolUse, bool isKnownModel, bool supportsBetweenToolsThinking = false)Creates a capability set.
Parameters
| Name | Type | Description |
|---|---|---|
maxOutputTokens |
int | |
supportsStructuredOutput |
bool | |
supportsAdaptiveThinking |
bool | |
rejectsSamplingParameters |
bool | |
supportsXhighEffort |
bool | |
rejectsThinkingDisabledAboveHighEffort |
bool | |
rejectsThinkingDisabled |
bool | |
rejectsBudgetThinking |
bool | |
rejectsForcedToolUse |
bool | |
isKnownModel |
bool | |
supportsBetweenToolsThinking |
bool |
Properties
Section titled “Properties”IsKnownModel
Section titled “IsKnownModel”public bool IsKnownModel { get; }The id is a recognized Claude model rather than a forward-compatible guess.
Value: bool
MaxOutputTokens
Section titled “MaxOutputTokens”public int MaxOutputTokens { get; }Model output token ceiling.
Value: int
RejectsBudgetThinking
Section titled “RejectsBudgetThinking”public bool RejectsBudgetThinking { get; }Budget-based thinking is rejected and sent as adaptive thinking.
Value: bool
RejectsForcedToolUse
Section titled “RejectsForcedToolUse”public bool RejectsForcedToolUse { get; }Forced tool choice is rejected.
Value: bool
RejectsSamplingParameters
Section titled “RejectsSamplingParameters”public bool RejectsSamplingParameters { get; }Temperature, top-p, and top-k are rejected.
Value: bool
RejectsThinkingDisabled
Section titled “RejectsThinkingDisabled”public bool RejectsThinkingDisabled { get; }The model always thinks and rejects disabled thinking.
Value: bool
RejectsThinkingDisabledAboveHighEffort
Section titled “RejectsThinkingDisabledAboveHighEffort”public bool RejectsThinkingDisabledAboveHighEffort { get; }Thinking cannot be disabled at effort above high.
Value: bool
SupportsAdaptiveThinking
Section titled “SupportsAdaptiveThinking”public bool SupportsAdaptiveThinking { get; }Thinking is adaptive rather than budget-based.
Value: bool
SupportsBetweenToolsThinking
Section titled “SupportsBetweenToolsThinking”public bool SupportsBetweenToolsThinking { get; }Thinking type between_tools is accepted. It is the lowest thinking setting on models that reject disabled thinking.
Value: bool
SupportsStructuredOutput
Section titled “SupportsStructuredOutput”public bool SupportsStructuredOutput { get; }Native output_config.format is available.
Value: bool
SupportsXhighEffort
Section titled “SupportsXhighEffort”public bool SupportsXhighEffort { get; }Effort xhigh is accepted.
Value: bool
Methods
Section titled “Methods”Get(string)
Section titled “Get(string)”public static AnthropicModelCapabilities Get(string modelId)Resolves capabilities from a model id, including Bedrock-style prefixes.
Parameters
| Name | Type | Description |
|---|---|---|
modelId |
string |
Returns: AnthropicModelCapabilities
Apache-2.0 · Copyright 2023 Vercel, Inc. · unofficial community port, not an official Vercel product