A Financial Times Study Backs Up What This Bittensor Subnet Has Been Selling All Along
Fine-tuned models beat general-purpose ones 85% to 50% on expert-level work, per FT reporting — which happens to be exactly the pitch behind Gradients' no-setup fine-tuning subnet.
BitExplorer · Jul 26, 2026
Gradients pointed to Financial Times reporting that general-purpose models — Gemini, Claude, GPT — scored a coin-flip 50% on expert-level work, while a version fine-tuned on the experts' own judgment hit 85%.
The argument, extending the FT piece: a written prompt can only convey the intuition an expert manages to put into words. A fine-tune captures more of what doesn't make it into the instructions.
That's directly the product Gradients sells — Rayon Labs' AutoML subnet for no-setup fine-tuning on Bittensor. Rayon Labs is also the team behind Chutes, currently the network's largest subnet by market cap.
Worth the caveat: this is Gradients amplifying third-party reporting that happens to support its own pitch, not an independent study of its own results.
Check the Gradients project page →
Source: @gradients_ai · Jul 20, 2026