Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Snowflake / Arctic managed task models

Arctic Transcribe

Managed speech-to-text and speaker-diarization models exposed through AI_TRANSCRIBE.

Transcription
availableClosed-source managed componentTranscription
Official documentation (opens in a new tab)Reviewed 15 Oct 2026
10 of 23 specification fields collected
Input content limit734M734,003,200
Languages32 listedFull language list in specifications
Input billingSee ratesSnowflake Cortex SQL
Output billingSee ratesPublished units and conditions below

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

Exact-model value not established by the reviewed service or research documentation.

Maximum input tokens
Not collected

Exact-model value not established by the reviewed service or research documentation.

Maximum output
Not collected

Exact-model value not established by the reviewed service or research documentation.

Parameters
Not collected

Exact-model value not established by the reviewed service or research documentation.

Reasoning
Not collected

Exact-model value not established by the reviewed service or research documentation.

Tool calling
Not collected

Exact-model value not established by the reviewed service or research documentation.

Structured output
Not collected

Exact-model value not established by the reviewed service or research documentation.

Streaming
Not collected

Exact-model value not established by the reviewed service or research documentation.

Batch processing
Not collected

Exact-model value not established by the reviewed service or research documentation.

Fine-tuning
Not collected

Exact-model value not established by the reviewed service or research documentation.

Context caching
Not collected

Exact-model value not established by the reviewed service or research documentation.

Release date
Not collected

Exact-model value not established by the reviewed service or research documentation.

Knowledge cutoff
Not collected

Exact-model value not established by the reviewed service or research documentation.

Provider model name
arctic-transcribeSource for Provider model name (opens in a new tab)

Function-specific model governance alias, not a versioned public checkpoint identifier.

Input modalities
Audio, VideoSource for Input modalities (opens in a new tab)
Output modalities
Text, JSONSource for Output modalities (opens in a new tab)
Audio duration limits
120 minutes without timestamps; 60 minutes with word or speaker timestampsSource for Audio duration limits (opens in a new tab)
Maximum input content size
734,003,200Source for Maximum input content size (opens in a new tab)

Documented 700 MB file limit and exact byte error threshold.

Input file formats
AAC, FLAC, M4A, MP3, MP4, OGG, WAV, WEBM, MKV, MOV, OGVSource for Input file formats (opens in a new tab)

Video must include an audio track. Allowed codecs: AAC, ALAC, FLAC, MP3, Opus, Vorbis and PCM.

Source audio handling
Internally resamples to 16 kHz mono. Language detection needs speech within the first five seconds.Source for Source audio handling (opens in a new tab)
Output formats
JSON transcript; optional word segments or speaker turns with start/end times.Source for Output formats (opens in a new tab)
Languages
View all 32 entries
  • Arabic
  • Bulgarian
  • Cantonese
  • Catalan
  • Chinese
  • Czech
  • Dutch
  • English
  • French
  • German
  • Greek
  • Hebrew
  • Hindi
  • Hungarian
  • Indonesian
  • Italian
  • Japanese
  • Korean
  • Latvian
  • Malay
  • Norwegian
  • Polish
  • Portuguese
  • Romanian
  • Russian
  • Serbian
  • Slovenian
  • Spanish
  • Swedish
  • Thai
  • Turkish
  • Ukrainian
Source for Languages (opens in a new tab)
Usage metering
50 billable tokens per second; minimum 10 seconds per file. One hour = 180,000 tokens.Source for Usage metering (opens in a new tab)

Access & pricing

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

Snowflake Cortex SQL

Provider source (opens in a new tab)
API model ID
Not collected

Implicit service routing; no versioned public checkpoint selector established.

Access
Requires SNOWFLAKE.CORTEX_USER plus the function/input object privileges. Model RBAC is checked first; CORTEX_MODELS_ALLOWLIST can permit the underlying model/alias as fallback.
Regions
AWS Oregon, N. Virginia and Frankfurt; Azure East US 2. Not available in PR China.
API endpoint URL
AI_TRANSCRIBE(audio_file [, options] [, return_error_details])Source for API endpoint URL (opens in a new tab)
Authentication
Model RBAC or an allowed underlying model/alias is required in addition to normal function/object privileges.Source for Authentication (opens in a new tab)
API parameters
options.timestamp_granularity can be word or speaker. Omit for an unsegmented transcript. return_error_details wraps value/error details.Source for API parameters (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Inference AI CreditsTranscription · AI Credits10.97 credits1M audio tokens

Rate conditions

  1. 50 audio tokens/second; minimum 10 seconds. The documentation dollar example assumes a particular credit price, which is not a universal USD tariff. Other platform charges may apply.
Pricing source (opens in a new tab)

Data & enterprise terms

Snowflake states that customer data is not used to train models made available to its customer base. Cross-region settings can permit transient payload processing outside the account region. Exact contractual retention terms are not established here.

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.

Snowflake Cortex SQL

Not verified

No exact-model prompt caching contract or discounted cache-token price established.

How it works
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

Minimum reusable input
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

Lifetime & refresh
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

How to enable it
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

Usage fields
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

Billing rules
Not collected

No exact-model prompt caching contract or discounted cache-token price established.

Limits & data boundaries
Published automatic response caching applies to text-only AI_COMPLETE, not this function. Its TTL and free hits are not assigned here.Source for Limits & data boundaries (opens in a new tab)

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

Caching reviewed 15 Oct 2026

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

A named Snowflake-managed function model alias. This service record does not establish a distinct private foundation checkpoint or disclose every constituent model.

Classification evidence (opens in a new tab)

Sources reviewed