Submit code models for evaluation on benchmarks