100B Daxiang Model Hits SOTA with Top Token Efficiency

What Changed

Model name: 大象 (Daxiang)

Why It Matters

This demonstrates efficient smaller models can match or exceed larger ones, slashing compute costs and enabling broader AI adoption in resource-constrained environments.

What To Do Next

Benchmark 大象 model on LMSYS arena to verify SOTA claims against GPT-4 class models.