A discussion on Hacker News explored claims that DeepSeek Online might be using hidden routing for its Fable 5 model. The analysis suggests that while observed response similarities might appear suspicious, they do not definitively prove hidden routing or distillation. The author emphasizes that such similarities can arise from common task setups, system instructions, or the inherent nature of good answers, rather than clandestine model usage. The piece advocates for rigorous testing by comparing explicitly chosen model IDs on specific tasks and preserving telemetry, rather than relying on anecdotal evidence like screenshots or matching response styles. AI
IMPACT Clarifies the difficulty in definitively proving hidden model routing, urging users to rely on rigorous testing rather than superficial similarities.
RANK_REASON The item discusses a claim made on Hacker News about a specific model's routing, analyzing the evidence and methodology for verification, rather than announcing a new release or product.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
