jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.6.1-cw-17K

TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Apr 8, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.6.1-cw-17K is a 7.6 billion parameter model fine-tuned from Snowflake/Arctic-Text2SQL-R1-7B. It specializes in Text-to-SQL generation for SQL++ queries, trained on the NL2SQL++ v8 dataset with code-with-thought reasoning. This model is optimized for accurately converting natural language questions into complex SQL++ queries, handling schema details and specific SQL++ syntax rules. It is particularly effective for analytical use cases requiring precise database interactions.

Loading preview...

Model Overview

This model, jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.6.1-cw-17K, is a 7.6 billion parameter variant of the Snowflake/Arctic-Text2SQL-R1-7B base model. It has been specifically fine-tuned for Text-to-SQL generation targeting SQL++ queries.

Key Capabilities

  • SQL++ Query Generation: Excels at converting natural language questions into valid and complex SQL++ queries, adhering to specific syntax and best practices (e.g., backtick enclosure for column names, 0-based indexing for SUBSTR, explicit column selection).
  • Schema Awareness: Utilizes a provided document schema to generate queries that accurately reflect the database structure, including handling nested objects and arrays.
  • Reasoning with Code-with-Thought: Fine-tuned on the NL2SQL++ v8 dataset, which incorporates "code-with-thought" reasoning, enabling the model to generate more robust and logically sound SQL queries by simulating a step-by-step thought process.
  • Complex Query Handling: Capable of generating queries involving UNNEST, JOIN operations, ARRAY_AGG, ROW_NUMBER() for ranking, and scalar subqueries, while adhering to strict rules against correlated subqueries or SELECT *.
  • Data Handling Nuances: Addresses specific requirements like IS NOT NULL checks for fields, IFMISSINGORNULL for aggregates, and proper handling of temporal filters and aggregations.

Good For

  • Automating SQL++ Query Writing: Ideal for developers and data analysts working with Snowflake's SQL++ who need to quickly generate complex queries from natural language.
  • Analytical Applications: Suitable for building applications that require dynamic SQL++ query generation for reporting, data exploration, and business intelligence against semi-structured data.
  • Educational Purposes: Can serve as a tool for understanding how natural language concepts map to intricate SQL++ constructs, especially with its "code-with-thought" training.

This model is particularly distinguished by its specialized focus on SQL++ and its training methodology that emphasizes logical reasoning for query construction.

Finetunes
2 models