4x B200

5 records across 2 models (quants: FP8, NVFP4).

hardwaremodelquantbackendtok/spp tok/sWttft sctx/tgdateclassflagssource
4x B200GLM-5.2-NVFP4 NVFP4vLLM132.5–––tg20482026-10-09community–source
4x B200GLM-5.2-NVFP4 NVFP4vLLM130.8–––tg20482026-10-09community–source
4x B200GLM-5.3-Flash FP8vLLM3332––––2026-10-09sourced–source
4x B200GLM-5.3-Flash FP8vLLM3152––––2026-10-09sourced–source
4x B200GLM-5.3-Flash FP8vLLM2063––––2026-10-09sourced–source

All rows cite their source. Provenance: community 40018 sourced 1085

← all hardware