Repo Prompt
Repo Prompt - A macOS native app designed to remove all the friction involved in iterating on your code with the most powerful AI models.
AI Model Benchmark
Summary
Brokk's Power Ranking is a benchmark comparing 30+ LLMs on real-world codebase tasks, measuring performance score, cost efficiency, and speed. Gemini 2.5 Pro leads at 75% performance, while GPT-5 High offers the best cost-performance ratio at 62% for $4.21.
Highlights
- Gemini 2.5 Pro achieves the highest performance score at 75%, followed by Claude 4.1 Opus at 72%
- GPT-5 High delivers strong cost efficiency with 62% performance at only $4.21 total spend
- GPT-OSS-120b is the cheapest option at $0.30 but scores only 39% on task completion
- Claude 4.1 Opus is fastest among top performers at 37 seconds total latency
- Performance degrades significantly for open-weight models like Kimi K2 (22%) and DeepSeek-R1 (31%)
auto-generated
Context
Audience
Software engineers and technical leads evaluating which LLM to use for AI-assisted coding on large or complex codebases
DomainLarge Language Models for Code
Formatinteractive benchmark dashboard
Accessfree online with optional paid credits for frontier models
Source Overview
Brokk is an AI-native code platform designed to provide compiler-grade context for large codebases, enabling accurate and efficient AI assistance across millions of lines of code.
It integrates advanced static analysis tools and supports deep scans, agentic search, multi-stage planning, and seamless handling of dependencies. Brokk facilitates direct supervision and collaboration between developers and AI, while maintaining transparency and control through an open-source client.
Topics
Related
LLM coding benchmarksCodebase-aware AI assistantsStatic analysis for LLM contextGPT-5Claude 4.1 OpusGemini 2.5 Pro
