The new DeepSeek-V4.1-Flash-1017 model delivers substantial improvements across benchmarks.
Alongside better performance, it reduces task runtime and output length, helping users finish work faster and at lower cost. It establishes a new performance–cost Pareto frontier across multiple harnesses. This release also focuses on frameworks such as DeepSeek Harness for a smoother experience when used with Tabbit.
DeepSeek-V4.1-Flash in Tabbit has been upgraded to the 1017 version. Click Switch Model at the bottom right of the chat box to try it.