Qwen3.5-397B
Qwen3.5-397B is a frontier Qwen large language model for reasoning, coding, search, and agent workflows. The underlying open-weight flagship uses a sparse MoE design with 397B total parameters and 17B activated parameters, supports 262K native context extensible to about 1M tokens, and is designed for high-throughput long-context inference.
API Options
Platform-level options for task execution and delivery.
-
taskType
string required value: textInference -
Identifier for the type of task being performed
-
taskUUID
string required UUID v4 -
UUID v4 identifier for tracking tasks and matching async responses. Must be unique per task.
-
webhookURL
string URI -
Specifies a webhook URL where JSON responses will be sent via HTTP POST when generation tasks complete. For batch requests with multiple results, each completed item triggers a separate webhook call as it becomes available.
Learn more 1 resource
- Webhooks PLATFORM
- Webhooks
-
deliveryMethod
string default: sync -
Determines how the API delivers task results.
Allowed values 3 values
- Returns complete results directly in the API response.
- Returns an immediate acknowledgment with the task UUID. Poll for results using getResponse.
- Streams results token-by-token as they are generated.
Learn more 1 resource
- Task Polling PLATFORM
-
includeCost
boolean default: false -
Include task cost in the response.
-
includeUsage
boolean default: false -
Include token usage statistics in the response.
-
numberResults
integer min: 1 max: 4 default: 1 -
Number of results to generate. Each result uses a different seed, producing variations of the same parameters.
Inputs
Input resources for the task (images, audio, etc). These must be nested inside the inputs object.
inputs object. Generation Parameters
Core parameters for controlling the generated content.
-
model
string required value: alibaba:qwen@3.5-397b -
Identifier of the model to use for generation.
Learn more 3 resources
-
seed
integer min: 0 max: 4294967295 -
Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range.
-
messages
array of objects required min items: 1 -
Array of chat messages forming the conversation context.
Settings
Technical parameters to fine-tune the inference process. These must be nested inside the settings object.
settings object.-
settings»systemPromptsystemPrompt
string min: 1 max: 262144 -
System-level instruction that guides the model's behavior and output style across the entire generation.
-
settings»temperaturetemperature
float min: 0 max: 2 step: 0.01 default: 0.6 -
Controls randomness in generation. Lower values produce more deterministic outputs, higher values increase variation and creativity.
-
settings»topPtopP
float min: 0 max: 1 step: 0.01 default: 0.95 -
Nucleus sampling parameter that controls diversity by limiting the probability mass. Lower values make outputs more focused, higher values increase diversity.
-
settings»frequencyPenaltyfrequencyPenalty
float min: 0 max: 2 step: 0.01 default: 0 -
Penalizes tokens based on their frequency in the output so far. A value of 0.0 disables the penalty.
-
settings»maxTokensmaxTokens
integer min: 1 max: 65536 -
Maximum number of tokens to generate in the response.
-
settings»minPminP
float min: 0 max: 1 step: 0.01 default: 0 -
Minimum probability threshold. Tokens with probability below this value are excluded from sampling.
-
settings»presencePenaltypresencePenalty
float min: 0 max: 2 step: 0.01 default: 0 -
Encourages the model to introduce new topics. A value of 0.0 disables the penalty.
-
settings»repetitionPenaltyrepetitionPenalty
float min: 0 max: 2 step: 0.01 default: 1 -
Penalizes tokens that have already appeared in the output. A value of 1.0 disables the penalty.
-
settings»stopSequencesstopSequences
array of strings min: 1 -
Array of sequences that will cause the model to stop generating further tokens when encountered.
-
settings»thinkingLevelthinkingLevel
string default: high -
Controls the depth of internal reasoning the model performs before generating a response.
Allowed values 2 values
-
settings»topKtopK
integer min: 1 max: 100 default: 20 -
Top-K sampling parameter that limits the number of highest-probability tokens considered at each step.