DeepSeek-V3.1 (Thinking) is an open source model from DeepSeek. It ranks #125 of 426 models on Text Arena · Instruction Following with a score of 1,416.
Rank spread 98 – 163.
76 points under the frontier
The median is $1.60.