AI Briefing10 AI Models Benchmarked on Bitcoin Price Prediction
Key point
In a 30-day Bitcoin prediction benchmark, Perplexity ranked first.
Details
Every day at 06:00 UTC, the same structured prompt was sent to 10 models to predict the Bitcoin price 7 days later, accumulating 25 data points per model over 30 days for scoring.
The score was calculated as 100 - min(100, abs(predicted - actual) / actual * 100). 100% means a perfect match, and 0% means an error of 100% or more.
The average accuracy was as follows.
- Perplexity: ranked 1st with 95.3%, best 100.0%
- Qwen and ChatGPT: 89.3% each
- DeepSeek: 85.8%
- Claude: 85.0%
- Grok: 77.6%
- Mistral: 63.7%, worst 34.5%, best 99.7%
- Llama: 59.1%
- Gemini 2.5 Flash: 12.2%, worst -43.0%
Since Perplexity is web-connected and can see the real-time Bitcoin price, it should be interpreted separately as to whether this is a simple prediction or a result of reading the current price and market signals.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Step 5 Preview Released: 600B Parameter MoE Model Highlights Agentic Tasks and Financial Analysis PerformanceQwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance Metrics