PulseAugur
EN
LIVE 03:46:53

Gemini 2.5 Flash beats Claude Sonnet 4.5 on Oracle Text-to-SQL benchmark

By PulseAugur Editorial · [1 sources] ·

A new benchmark called UniQL, designed to translate natural language questions into SQL queries across various dialects, has highlighted the performance differences between large language models. On the Oracle dialect of UniQL, Claude Sonnet 4.5 achieved 49.61% accuracy, while Gemini 2.5 Pro scored 26.14%. By implementing specific prompt rules tailored to Oracle's syntax and reserved words, the Gemini 2.5 Flash model was able to improve its single-try accuracy from 35% to 64.5%, surpassing the previously best-reported score. AI

IMPACT Demonstrates the significant impact of prompt engineering and database-specific rules on LLM performance for Text-to-SQL tasks.

RANK_REASON New benchmark and model performance comparison on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Gemini 2.5 Flash beats Claude Sonnet 4.5 on Oracle Text-to-SQL benchmark

How we ranked this

Signal score
49 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
New benchmark and model performance comparison on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
model release, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Ashish sinha ·

    Text-to-SQL on Oracle: two prompt rules took a light model from 35% to 64%

    <p><a href="https://github.com/JerryGao818/UniQL" rel="noopener noreferrer">UniQL</a> is a new benchmark that translates BIRD's 1,534 text-to-SQL questions into 16 SQL dialects. Its Oracle column is hard. The best published result is <strong>49.61%</strong> (Claude Sonnet 4.5), a…

RELATED ENTITIES

RELATED TOPICS