From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix
Open paper
Benchmark
Model
Rank
Results
agents-on-bench
T-pro2.1 (internal, non-reasoning)
–
Accuracy: 37.6
instruction-following-on-multichallenge
T-pro-2.1 public
–
Accuracy: 37.4
Rank counts only results with a code link.