AI Model Arena & Latency Benchmarks
Compare top LLMs side-by-side with blind evaluation and real-time inference speed tracking.
- Open https://wilbot.link/arena in any modern browser.
Architecture & Overview
The WilBot AI Model Arena (https://wilbot.link/arena) is an open LLM benchmarking arena. Users submit prompts that are simultaneously dispatched to two competing state-of-the-art models.
To eliminate cognitive bias, model identities are masked as 'Model Alpha' and 'Model Beta' until the user votes on the superior response. The arena tracks inference latency in milliseconds, win-rate percentages, and updates global Elo ratings.
Key Features & Capabilities
Step-by-Step Walkthrough
Navigate to the Arena
Visit https://wilbot.link/arena.
https://wilbot.link/arenaSubmit a Benchmark Prompt
Enter a challenging coding problem, logic riddle, or creative writing prompt, then click 'Launch Battle'.
Inspect Outputs & Cast Your Vote
Review both answers and click 'Vote Alpha' or 'Vote Beta'. Once submitted, the true model identities are revealed alongside their latency numbers.
Real-World Examples & Outputs
Live Arena API Endpoint
Developers can programmatically run arena battles.POST https://wilbot.link/api/arenaTroubleshooting & Common Pitfalls
Frequently Asked Questions
Which models compete in the Arena?
The arena rotates top models including Gemini 3.6 Flash, Llama 3.3 70B (Groq), Qwen 2.5 72B (OpenRouter), and Llama 3.1 8B (Cerebras).