Skip to content
Model infoModelsSnapshot: 15 Oct 2026

iFLYTEK / Spark

Spark Max 32K

Native Spark text model with separate HTTP and WebSocket access.

Language models (LLMs)
availableClosed weightsText generation
Official documentation (opens in a new tab)Reviewed 12 Oct 2026
5 of 17 specification fields collected
Input token limit32KSee specification notes below
Maximum outputNot collectedAwaiting source review
Input billingNot collectediFLYTEK
Output billingNot collectedNo inferred price

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum output
Not collected

Conflicting limits: overview says 8K output; HTTP/WebSocket max_tokens allows 32,768.

Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Reasoning
Not collected
Tool calling
Not collected

Function calling is described for Max/Ultra, but the exact Max-32K selector is not explicitly confirmed.

Structured output
SupportedSource for Structured output (opens in a new tab)

JSON-object mode is documented. This does not establish strict JSON Schema enforcement; the prompt must ask for JSON.

Streaming
SupportedSource for Streaming (opens in a new tab)
Batch processing
Not collected
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
Not collected
Architecture
Not collected
Languages
Not collected
Context caching
Not collected
Maximum input tokens
32KSource for Maximum input tokens (opens in a new tab)

Provider uses K notation; an exact decimal/binary expansion is not asserted.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

iFLYTEK

Provider source (opens in a new tab)
API model ID
max-32k
Access
Authenticated native service. Model authorization and the corresponding API credentials are required. No paid inference was performed.
Regions
China-hosted service; endpoint hostname is recorded, but no customer-specific residency or international eligibility guarantee is inferred.
API endpoint URL
POST https://spark-api-open.xf-yun.com/v1/chat/completionsSource for API endpoint URL (opens in a new tab)
Maximum output
32,768 tokensSource for Maximum output (opens in a new tab)

Parameter upper bound; not proof that the lower overview limit is obsolete.

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

The LLM API service agreement reserves provider IP and keeps input/output ownership with the user or original rights holder. The consumer SparkDesk privacy policy is not treated as an API no-training or zero-retention contract.

Read the service policy (opens in a new tab)

iFLYTEK

Provider source (opens in a new tab)
API model ID
max-32k
Access
Authenticated native service. Model authorization and the corresponding API credentials are required. No paid inference was performed.
Regions
China-hosted service; endpoint hostname is recorded, but no customer-specific residency or international eligibility guarantee is inferred.
API endpoint URL
wss://spark-api.xf-yun.com/chat/max-32kSource for API endpoint URL (opens in a new tab)
Tool calling
Not supportedSource for Tool calling (opens in a new tab)

The WebSocket guide directs function-calling use to HTTP.

Maximum output
32,768 tokensSource for Maximum output (opens in a new tab)

Parameter upper bound; overview conflicts are preserved.

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

The LLM API service agreement reserves provider IP and keeps input/output ownership with the user or original rights holder. The consumer SparkDesk privacy policy is not treated as an API no-training or zero-retention contract.

Read the service policy (opens in a new tab)

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

iFLYTEK

max-32k
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

iFLYTEK

max-32k
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Inferred closed-weight native Spark service: the provider documents this named service behind authenticated API access and reserves service IP in its LLM API terms. This is distinct from the explicitly released XHToken 4B/1.7B weights and the third-party open models hosted by MaaS. API or SDK source availability is not model-weight availability.

Classification evidence (opens in a new tab)

Sources reviewed