List your company

Structured Generation on C++

1.42Mean Generation Latency (s)Llama-3B
1.27122.27563.284.2844May 28, 2026

Evaluation results

MethodLinks
Llama-3BConstraint=Unconstrained2026.051.42
Llama-3BConstraint=+CFGZIP2026.051.47
Llama-3BConstraint=XGrammar22026.052.14
GPT-20BConstraint=Unconstrained2026.052.34
GPT-20BConstraint=+CFGZIP2026.052.41
Qwen-4BConstraint=Unconstrained2026.052.48
Qwen-4BConstraint=+CFGZIP2026.052.55
GPT-20BConstraint=XGrammar22026.054.59
Qwen-4BConstraint=XGrammar22026.055.14