AnalysisAI ModelsSeptember 16, 2026

Qwen 3.8 27B runs 63 hours on an RTX 3090 on the Riemann hypothesis

A 4-bit quantized Qwen 3.8 27B with a 100K context window ran autonomously for 63 hours, consuming over 50 million tokens, without solving the Riemann hypothesis. The poster says the run still reveals the model's internal work, memory organization, and strategies.

1 source

More stories today

Open the live feed