Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
To make LLMs better, we need to feed them the best Solana code possible What is your favorite repo example of good Solana code?
It's incredibly important that LLMs get good at building transactions on Solana Solana Bench helps us understand where LLMs can improve and which is the best Continuous improvement is how we'll get LLMs better and accelerate development S/O to @ngundotra for the amazing work https://t.co/QFWVPPWBYF
🚨NEW: @SolanaFndn launches Solana Bench, an open-source benchmark that evaluates how AI models (Claude, GPT-5, Gemini, etc.) can generate and execute transactions with @Solana SDKs and DeFi protocols, alongside new funding for research. https://t.co/yH50oBM8LF
The Solana Foundation has released Solana Bench, an open-source benchmark designed to measure how effectively large language models such as Claude, GPT-5 and Gemini can generate and execute on-chain transactions using Solana software-development kits and DeFi protocols. Foundation officials said the tool will give developers a standardized way to compare model performance, identify weaknesses and prioritize improvements, with the goal of automating more blockchain tasks and broadening access to Solana-based applications. The organization also committed additional, undisclosed funding to support further research on AI-driven transaction building.