Model
57d agoBridgewater Fine-Tunes Qwen For Finance
An 84.7 percent accuracy rate is the claimed result of a fine-tuned Qwen3-235B model developed by Bridgewater and Thinking Machines Lab. The system reportedly beats GPT and Claude at one-fourteenth of the cost. These figures remain unverified by third parties.
The Signal
The result suggests specialized fine-tuning outperforms general-purpose LLMs on private financial data.