ssakethch/h200-quantization-benchmarks

H200 Quantization Benchmarks Benchmark results for 40 quantized and non-quantized instruction-tuned LLMs evaluated on an NVIDIA H200 MIG (Multi-Instance GPU) setup. This dataset supports reproducible comparison of quantization methods (AWQ, GPTQ, fp8, bf16) across accuracy and throughput dimensions. Dataset Configs Config Description Rows accuracy Per-task accuracy results from lm-eval across all models ~240 accuracy_leaderboard Aggregated accuracy… See the full description on the dataset page: https://huggingface.co/datasets/ssakethch/h200-quantization-benchmarks.