AISQL functions¶

Task-specific functions are purpose-built and managed functions that automate routine tasks, like simple summaries andquick translations, that don’t require any customization.

• AI_COMPLETE: Generates a completion for a given text string or image using a selected LLM. Use this function for most generative AI tasks.

  • This is the updated version ofCOMPLETE (SNOWFLAKE.CORTEX).

• AI_CLASSIFY: Classifies text or images into user-defined categories.

  • This is the updated version ofCLASSIFY_TEXT (SNOWFLAKE.CORTEX) with support for multi-label and image classification.

• AI_FILTER: Returns True or False for a given text or image input, allowing you to filter results inSELECT,WHERE, orJOIN...ON clauses.

• AI_AGG: Aggregates a text column and returns insights across multiple rows based on a user-defined prompt. This function isn’t subject to context window limitations.

• AI_EMBED: Generates an embedding vector for a text or image input, which can be used for similarity search, clustering, and classification tasks.

  • This is the updated version ofEMBED_TEXT_1024 (SNOWFLAKE.CORTEX).

• AI_SUMMARIZE_AGG: Aggregates a text column and returns a summary across multiple rows. This function isn’t subject to context window limitations.

• AI_SIMILARITY: Calculates the embedding similarity between two inputs.

• PARSE_DOCUMENT (SNOWFLAKE.CORTEX): Extracts text (using OCR mode) or text with layout information (using LAYOUT mode) from documents in an internal or external stage.

• TRANSLATE (SNOWFLAKE.CORTEX): Translates text between supported languages.

• SENTIMENT (SNOWFLAKE.CORTEX): Extracts sentiment scores from text.

• EXTRACT_ANSWER (SNOWFLAKE.CORTEX): Extracts the answer to a question from unstructured data, provided that the relevant data exists.

• SUMMARIZE (SNOWFLAKE.CORTEX): Returns a summary of the text that you’ve specified.

Helper functions¶

Helper functions are purpose-built and managed functions that reduce cases of failures when running other AISQL functions, for example bygetting the count of tokens in an input prompt to ensure the call doesn’t exceed a model limit.

• COUNT_TOKENS (SNOWFLAKE.CORTEX): Given an input text, returns the token count based on the model or Cortexfunction specified.

• TRY_COMPLETE (SNOWFLAKE.CORTEX): Works like the COMPLETE function, but returns NULLwhen the function could not execute instead of an error code.

Cortex Guard¶

Cortex Guard is an option of the COMPLETE function designed to filter possible unsafe and harmful responses from alanguage model. Cortex Guard is currently built with Meta’s Llama Guard 3. Cortex Guard works by evaluating the responses of a languagemodel before that output is returned to the application. Once you activate Cortex Guard, language model responses which may be associatedwith violent crimes, hate, sexual content, self-harm, and more are automatically filtered. SeeCOMPLETE arguments for syntax and examples.

Model allowlist¶

Use theCORTEX_MODELS_ALLOWLIST parameter in theALTERACCOUNTSET command to set model access for all users in the account.If you need to provide specific users with access beyond what you’ve specified in the allowlist, you should use role-based access control instead.For more information about role-based access control, seeRole based access control.

When your users make a request, Snowflake Cortex evaluates the parameter to determine whether the user can access the model.

For theCORTEX_MODELS_ALLOWLIST parameter, you can set the following values:

• CORTEX_MODELS_ALLOWLIST='All'Provides access to all models.

  The following command provides user access to all models:

  ALTERACCOUNTSETCORTEX_MODELS_ALLOWLIST='All';

  Copy

• CORTEX_MODELS_ALLOWLIST='model1,model2,...'

  Provides users with access to the models specified in a comma-separated list.

  The following command provides users with access to themistral-large2 andllama3.1-70b models:

  ALTERACCOUNTSETCORTEX_MODELS_ALLOWLIST='mistral-large2,llama3.1-70b';

  Copy

• CORTEX_MODELS_ALLOWLIST='None'Prevents users from accessing any model.

  The following command prevents user access to any model:

  ALTERACCOUNTSETCORTEX_MODELS_ALLOWLIST='None';

  Copy

Role based access control¶

Each model in Snowflake Cortex is a unique object in the SNOWFLAKE.MODELS schema with an associated application role.You can use the model objects and application roles to manage access to the model object.

As a role with the ACCOUNTADMIN privilege, run the following command to get access to the latest models:

CALLSNOWFLAKE.MODELS.CORTEX_BASE_MODELS_REFRESH();

Next, use the following command to list the models that are available for your current role:

SHOWMODELSINSNOWFLAKE.MODELS;

The command returns a list of models, such as the following:

Use the following command to list the application roles for these models:

SHOWAPPLICATIONROLESINAPPLICATIONSNOWFLAKE;

The command returns a list of application roles, such as the following:

To grant access to a specific model, you grant the model’s application role to a user role.For example, you can grantCORTEX-MODEL-ROLE-LLAMA3.1-70B, the application role forSNOWFLAKE.MODELS."LLAMA3.1-70B", to a user role.The following command grants theCORTEX-MODEL-ROLE-LLAMA3.1-70B application role to theMY_ROLE user role:

GRANTAPPLICATION ROLESNOWFLAKE."CORTEX-MODEL-ROLE-LLAMA3.1-70B"TOROLEMY_ROLE;

To make an inference call, use the fully qualified model name.The following is an example of a call users can make:

SELECTAI_COMPLETE('SNOWFLAKE.MODELS."LLAMA3.1-70B"','Hello');

Track costs for AI services¶

To track credits used for AI Services including LLM Functions in your account, use theMETERING_HISTORY view:

SELECT*FROMSNOWFLAKE.ACCOUNT_USAGE.METERING_DAILY_HISTORYWHERESERVICE_TYPE='AI_SERVICES';

Track credit consumption for AISQL functions¶

To view the credit and token consumption for each AISQL function call, use theCORTEX_FUNCTIONS_USAGE_HISTORY view:

SELECT*FROMSNOWFLAKE.ACCOUNT_USAGE.CORTEX_FUNCTIONS_USAGE_HISTORY;

You can also view the credit and token consumption for each query within your Snowflake account. Viewing the credit and token consumption for each query helps you identify queries that are consuming the most credits and tokens.

The following example query uses theCORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view to show the credit and token consumption for all of your queries within your account.

SELECT*FROMSNOWFLAKE.ACCOUNT_USAGE.CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY;

You can also use the same view to see the credit and token consumption for a specific query.

SELECT*FROMSNOWFLAKE.ACCOUNT_USAGE.CORTEX_FUNCTIONS_QUERY_USAGE_HISTORYWHEREquery_id='<query-id>';

The query usage history is grouped by the models used in the query. For example, if you ran:

SELECTAI_COMPLETE('mistral-7b','Is a hot dog a sandwich'),AI_COMPLETE('mistral-large','Is a hot dog a sandwich');

The query usage history would show two rows, one formistral-7b and one formistral-large.

Large models¶

If you’re not sure where to start, try the most capable models first to establish a baseline to evaluate other models.claude-3-7-sonnet,reka-core, andmistral-large2 are the most capable models offered by Snowflake Cortex,and will give you a good idea what a state-of-the-art model can do.

• Claude3-7Sonnet is a leader in general reasoning and multimodal capabilities. It outperforms its predecessors in tasks that require reasoning across different domains and modalities. You can use its large output capacity to get more information from either structured or unstructured queries. Its reasoning capabilities and large context windows make it well-suited for agentic workflows.

• deepseek-r1 is a foundation model trained using large-scale reinforcement-learning (RL) without supervised fine-tuning (SFT).It can deliver high performance across math, code, and reasoning tasks.To access the model, set thecross-region inference parameter toAWS_US.

• mistral-large2 is Mistral AI’s most advanced large language model with top-tier reasoning capabilities.Compared tomistral-large, it’s significantly more capable in code generation, mathematics, reasoning, andprovides much stronger multilingual support. It’s ideal for complex tasks that require large reasoning capabilitiesor are highly specialized, such as synthetic text generation, code generation, and multilingual text analytics.

• llama3.1-405b is an open source model from thellama3.1 model family from Meta with a large 128K context window.It excels in long document processing, multilingual support, synthetic data generation and model distillation.

• snowflake-llama3.1-405b is a model derived from the open source llama3.1 model. It uses the<SwiftKVoptimizationshttps://www.snowflake.com/en/blog/up-to-75-lower-inference-cost-llama-meta-llm/> that have been developed by the Snowflake AI research team to deliver up to a 75% inference cost reduction. SwiftKV achieves higher throughput performance with minimal accuracy loss.

Medium models¶

• llama3.1-70b is an open source model that demonstrates state-of-the-art performance ideal for chat applications,content creation, and enterprise applications. It is a highly performant, cost effective model that enables diverse usecases with a context window of 128K.llama3-70b is still supported and has a context window of 8K.

• snowflake-llama3.3-70b is a model derived from the open source llama3.3 model. It uses the<SwiftKVoptimizationshttps://www.snowflake.com/en/blog/up-to-75-lower-inference-cost-llama-meta-llm/> developed by the Snowflake AI research team to deliver up to a 75% inference cost reduction. SwiftKV achieves higher throughput performance with minimal accuracy loss.

• snowflake-arctic is Snowflake’s top-tier enterprise-focused LLM. Arctic excels at enterprise tasks such as SQLgeneration, coding and instruction following benchmarks.

• mixtral-8x7b is ideal for text generation, classification, and question answering. Mistral models are optimizedfor low latency with low memory requirements, which translates into higher throughput for enterprise use cases.

• Thejamba-Instruct model is built by AI21 Labs to efficiently meet enterprise requirements. It is optimized to offer a 256k tokencontext window with low cost and latency, making it ideal for tasks like summarization, Q&A, and entity extraction on lengthy documentsand extensive knowledge bases.

• The AI21 Jamba 1.5 family of models is state-of-the-art, hybrid SSM-Transformer instruction following foundation models. Thejamba-1.5-mini andjamba-1.5-large with a context length of 256K supports use cases such as structured output (JSON), and grounded generation.

Small models¶

• Thellama3.2-1b andllama3.2-3b models support context length of 128K tokens and are state-of-the-art in their class for usecases like summarization, instruction following, and rewriting tasks. The Llama 3.2 models deliver multilingual capabilities, withsupport for English, German, French, Italian, Portuguese, Hindi, Spanish and Thai.

• llama3.1-8b is ideal for tasks that require low to moderate reasoning. It’s a light-weight, ultra-fast model with a context windowof 128K.llama3-8b andllama2-70b-chat are still supported models that provide a smaller context window and relatively lower accuracy.

• mistral-7b is ideal for your simplest summarization, structuration, and question answering tasks that need to bedone quickly. It offers low latency and high throughput processing for multiple pages of text with its 32K contextwindow.

• gemma-7b is suitable for simple code and text completion tasks. It has a context window of 8,000 tokens but issurprisingly capable within that limit, and quite cost-effective.

The following table provides information on how popular models perform on various benchmarks,including the models offered by Snowflake Cortex COMPLETE as well as a few other popular models.