{"generated_at":"2026-10-31T00:02:30Z","normalization":"percentages are relative to the best reported recipe on the same GPU generation; speeds are per GPU (tok/s divided by device_count); median across recorded points, not max","weights":{"decode":0.5500000000000000444089209850062616169452667236328125,"prefill":0.1499999999999999944488848768742172978818416595458984375,"quality":0.299999999999999988897769753748434595763683319091796875,"fallback_without_eval":{"decode":0.8000000000000000444089209850062616169452667236328125,"prefill":0.200000000000000011102230246251565404236316680908203125}},"meta":{"count":74,"origin":"https://localairecipes.com","openapi":"/openapi.json"},"recipes":[{"slug":"deepseek-v4-1-flash-1017-fp8-tp1xpp2-dspark-2x-dgx-station-gb300","title":"DeepSeek V4.1 Flash 1017 FP8 TP1xPP2 DSpark 2x DGX Station GB300 RoCE vLLM 0.31 overlay","tier":"community","url":"/r/deepseek-v4-1-flash-1017-fp8-tp1xpp2-dspark-2x-dgx-station-gb300","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-station-gb300","device_count":2,"vram_gb":288},"score":80,"score_basis":"speed-only","decode":{"tok_s":357,"pct_of_generation_best":100},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-1-flash-4x6000-jj-dspark-k7","title":"DeepSeek V4.1 Flash on 4x RTX PRO 6000 Blackwell \\x{2014} TP4 DSpark K7","tier":"community","url":"/r/ds4-1-flash-4x6000-jj-dspark-k7","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":4,"vram_gb":96},"score":71.7000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":230.599999999999994315658113919198513031005859375,"pct_of_generation_best":64.599999999999994315658113919198513031005859375},"prefill":{"tok_s":17503,"pct_of_generation_best":100},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-4x6000-jj-mtp3","title":"GLM-5.3-Flash on 4x RTX PRO 6000 Blackwell \\x{2014} TP4 MTP3","tier":"community","url":"/r/glm-5-3-flash-4x6000-jj-mtp3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":4,"vram_gb":96},"score":71.2000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":248.900000000000005684341886080801486968994140625,"pct_of_generation_best":69.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":13514,"pct_of_generation_best":77.2000000000000028421709430404007434844970703125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-4x6000-kk-dflash2-k7","title":"GLM-5.3-Flash on 4x RTX PRO 6000 Blackwell \u2014 Karmic Kraken TP4 DFlash2 K7","tier":"community","url":"/r/glm-5-3-flash-4x6000-kk-dflash2-k7","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":4,"vram_gb":96},"score":67.2999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":229,"pct_of_generation_best":64.099999999999994315658113919198513031005859375},"prefill":{"tok_s":13955,"pct_of_generation_best":79.7000000000000028421709430404007434844970703125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-rtx-pro-6000-blackwell-sm120-sglang","title":"Qwen3.8-Flash-Next NVFP4 RTX PRO 6000 Blackwell SM120 SGLang","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-rtx-pro-6000-blackwell-sm120-sglang","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":65.599999999999994315658113919198513031005859375,"score_basis":"speed-only","decode":{"tok_s":224.099999999999994315658113919198513031005859375,"pct_of_generation_best":62.7999999999999971578290569595992565155029296875},"prefill":{"tok_s":13460.70000000000072759576141834259033203125,"pct_of_generation_best":76.900000000000005684341886080801486968994140625},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-4x6000-jj-dflash2-k7","title":"GLM-5.3-Flash on 4x RTX PRO 6000 Blackwell \\x{2014} TP4 DFlash2 K7","tier":"community","url":"/r/glm-5-3-flash-4x6000-jj-dflash2-k7","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":4,"vram_gb":96},"score":65,"score_basis":"speed-only","decode":{"tok_s":219.69999999999998863131622783839702606201171875,"pct_of_generation_best":61.5},"prefill":{"tok_s":13840,"pct_of_generation_best":79.099999999999994315658113919198513031005859375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-kraken-tp3-3x-rtx-pro-6000-blackwell","title":"DeepSeek-V4.1-Flash-1017 Kraken TP3 3x RTX PRO 6000 Blackwell","tier":"community","url":"/r/deepseek-v4-1-flash-1017-kraken-tp3-3x-rtx-pro-6000-blackwell","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":3,"vram_gb":96},"score":64.2999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":228.5,"pct_of_generation_best":64},"prefill":{"tok_s":11445.5,"pct_of_generation_best":65.400000000000005684341886080801486968994140625},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-flash-2x6000-jj-dspark-k5","title":"DeepSeek V4 Flash-0731 on 2x RTX PRO 6000 Blackwell \\x{2014} TP2 DSpark K5","tier":"community","url":"/r/ds4-flash-2x6000-jj-dspark-k5","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":2,"vram_gb":96},"score":60.89999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":214.900000000000005684341886080801486968994140625,"pct_of_generation_best":60.2000000000000028421709430404007434844970703125},"prefill":{"tok_s":11167,"pct_of_generation_best":63.7999999999999971578290569595992565155029296875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-1x6000-kk-mtp3","title":"Qwen3.8-Flash-Next on RTX PRO 6000 Blackwell \u2014 Karmic Kraken TP1 MTP3","tier":"community","url":"/r/qwen3-8-flash-next-1x6000-kk-mtp3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":52.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":173.19999999999998863131622783839702606201171875,"pct_of_generation_best":48.5},"prefill":{"tok_s":12073,"pct_of_generation_best":69},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-rtx-pro-6000-blackwell-sglang-nvme-ple","title":"Qwen3.8-Flash-Next NVFP4 RTX PRO 6000 Blackwell SGLang NVMe PLE","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-rtx-pro-6000-blackwell-sglang-nvme-ple","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":51.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":180.900000000000005684341886080801486968994140625,"pct_of_generation_best":50.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":9692.5,"pct_of_generation_best":55.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-vision-2x6000-jj-dspark-k3","title":"DeepSeek V4 Flash Vision on 2x RTX PRO 6000 Blackwell \\x{2014} TP2 DSpark K3","tier":"community","url":"/r/ds4-vision-2x6000-jj-dspark-k3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":2,"vram_gb":96},"score":48.2000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":169.19999999999998863131622783839702606201171875,"pct_of_generation_best":47.39999999999999857891452847979962825775146484375},"prefill":{"tok_s":8993,"pct_of_generation_best":51.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-qad-csf-on-2x-rtx-pro-6000-blackwell-tp2-dcp2-mtp3","title":"GLM-5.3-Flash QAD (CSF) on 2x RTX PRO 6000 Blackwell \u2014 TP2/DCP2 MTP3","tier":"community","url":"/r/glm-5-3-flash-qad-csf-on-2x-rtx-pro-6000-blackwell-tp2-dcp2-mtp3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":2,"vram_gb":96},"score":46.39999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":207,"pct_of_generation_best":58},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-1x6000-jj-mtp3","title":"Qwen3.8-Flash-Next NVFP4 on RTX PRO 6000 Blackwell \u2014 TP1 MTP3","tier":"curated","url":"/r/qwen3-8-flash-next-nvfp4-1x6000-jj-mtp3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":38.2999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":137.19999999999998863131622783839702606201171875,"pct_of_generation_best":38.39999999999999857891452847979962825775146484375},"prefill":{"tok_s":6627.3000000000001818989403545856475830078125,"pct_of_generation_best":37.89999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-nvfp4a4-awq-1x-rtx-pro-6000-sglang-fast-decode","title":"Qwen3.8-27B NVFP4A4-AWQ 1x RTX PRO 6000 SGLang Fast Decode","tier":"community","url":"/r/qwen3-8-27b-nvfp4a4-awq-1x-rtx-pro-6000-sglang-fast-decode","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":29.5,"score_basis":"speed-only","decode":{"tok_s":131.69999999999998863131622783839702606201171875,"pct_of_generation_best":36.89999999999999857891452847979962825775146484375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-on-4x-dgx-spark-tuned-tp4-sglang-dspark","title":"DeepSeek-V4.1-Flash on 4x DGX Spark \u2014 tuned TP4 SGLang/DSpark","tier":"community","url":"/r/deepseek-v4-1-flash-on-4x-dgx-spark-tuned-tp4-sglang-dspark","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":4,"vram_gb":128},"score":26.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":89.7000000000000028421709430404007434844970703125,"pct_of_generation_best":25.10000000000000142108547152020037174224853515625},"prefill":{"tok_s":5650,"pct_of_generation_best":32.2999999999999971578290569595992565155029296875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-dgx-spark-gb10-sglang-nextn-mtp","title":"Qwen3.8-Flash-Next NVFP4 DGX Spark GB10 SGLang NEXTN MTP","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-dgx-spark-gb10-sglang-nextn-mtp","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":24.699999999999999289457264239899814128875732421875,"score_basis":"speed-only","decode":{"tok_s":110.400000000000005684341886080801486968994140625,"pct_of_generation_best":30.89999999999999857891452847979962825775146484375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exlr8-exl3-experts-on-dgx-spark-tp1-4","title":"GLM-5.3-Flash exlr8 EXL3 experts on DGX Spark TP1-4","tier":"community","url":"/r/glm-5-3-flash-exlr8-exl3-experts-on-dgx-spark-tp1-4","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":4,"vram_gb":128},"score":18.5,"score_basis":"speed-only","decode":{"tok_s":65.2000000000000028421709430404007434844970703125,"pct_of_generation_best":18.300000000000000710542735760100185871124267578125},"prefill":{"tok_s":3438.5,"pct_of_generation_best":19.60000000000000142108547152020037174224853515625},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exl3-3-05bpw-on-1x-rtx-pro-6000-gpu-expert-cache-h","title":"GLM-5.3-Flash EXL3 3.05bpw on 1x RTX PRO 6000 - GPU expert cache, host-RAM streaming","tier":"community","url":"/r/glm-5-3-flash-exl3-3-05bpw-on-1x-rtx-pro-6000-gpu-expert-cache-h","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":17.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":64.7999999999999971578290569595992565155029296875,"pct_of_generation_best":18.199999999999999289457264239899814128875732421875},"prefill":{"tok_s":2736.09999999999990905052982270717620849609375,"pct_of_generation_best":15.5999999999999996447286321199499070644378662109375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-bit-exact-spec-decode-1x-rtx-pro-6000-sglang","title":"Qwen3.8-27B Bit-Exact Spec-Decode 1x RTX PRO 6000 SGLang","tier":"community","url":"/r/qwen3-8-27b-bit-exact-spec-decode-1x-rtx-pro-6000-sglang","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":1,"vram_gb":96},"score":16.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":73.900000000000005684341886080801486968994140625,"pct_of_generation_best":20.699999999999999289457264239899814128875732421875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exl3-2-05bpw-on-2x-rtx-5090-tabbyapi-hot-expert-sw","title":"GLM-5.3-Flash EXL3 2.05bpw on 2x RTX 5090 - TabbyAPI hot-expert swaps, 262K ctx, vision","tier":"community","url":"/r/glm-5-3-flash-exl3-2-05bpw-on-2x-rtx-5090-tabbyapi-hot-expert-sw","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"rtx-5090","device_count":2,"vram_gb":32},"score":16.39999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":62.89999999999999857891452847979962825775146484375,"pct_of_generation_best":17.60000000000000142108547152020037174224853515625},"prefill":{"tok_s":2023,"pct_of_generation_best":11.5999999999999996447286321199499070644378662109375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-1x-dgx-spark-vllm-71-3-c1-salient-data","title":"Qwen3.8-Flash-Next NVFP4 1x DGX Spark vLLM 71.3 c1 (salient-data)","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-1x-dgx-spark-vllm-71-3-c1-salient-data","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":15.5,"score_basis":"speed-only","decode":{"tok_s":69.2999999999999971578290569595992565155029296875,"pct_of_generation_best":19.39999999999999857891452847979962825775146484375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-sglang-nvfp4-dflash2-on-1x-dgx-spark-deterministic","title":"Qwen3.8-27B SGLang NVFP4 DFlash2 on 1x DGX Spark (deterministic)","tier":"community","url":"/r/qwen3-8-27b-sglang-nvfp4-dflash2-on-1x-dgx-spark-deterministic","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":15.300000000000000710542735760100185871124267578125,"score_basis":"speed-only","decode":{"tok_s":68.2999999999999971578290569595992565155029296875,"pct_of_generation_best":19.10000000000000142108547152020037174224853515625},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-fp4-local-1x-dgx-spark-vllm-b12x","title":"Qwen3.8-Flash-Next FP4-Local 1x DGX Spark vLLM b12x","tier":"community","url":"/r/qwen3-8-flash-next-fp4-local-1x-dgx-spark-vllm-b12x","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":14.800000000000000710542735760100185871124267578125,"score_basis":"speed-only","decode":{"tok_s":66.2000000000000028421709430404007434844970703125,"pct_of_generation_best":18.5},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"nemotron-3-5-bolt-74b-a9b-1x-dgx-spark-vllm-b12x-lilicorr-262k","title":"Nemotron 3.5 Bolt 74B-A9B 1x DGX Spark vLLM b12x LiLiCorr 262K","tier":"community","url":"/r/nemotron-3-5-bolt-74b-a9b-1x-dgx-spark-vllm-b12x-lilicorr-262k","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":14.0999999999999996447286321199499070644378662109375,"score_basis":"speed-only","decode":{"tok_s":63,"pct_of_generation_best":17.60000000000000142108547152020037174224853515625},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-autoround-dgx-spark-vllm-mtp4-iter7g","title":"Qwen3.8-Flash-Next W4A16-AutoRound DGX Spark vLLM MTP4 Iter7g","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-autoround-dgx-spark-vllm-mtp4-iter7g","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":13.699999999999999289457264239899814128875732421875,"score_basis":"speed-only","decode":{"tok_s":61.2999999999999971578290569595992565155029296875,"pct_of_generation_best":17.199999999999999289457264239899814128875732421875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-uncensored-nvfp4-2x-dgx-spark-sglang-tp2-roce","title":"Qwen3.8-Flash-Next Uncensored NVFP4 2x DGX Spark SGLang TP2 ROCE","tier":"community","url":"/r/qwen3-8-flash-next-uncensored-nvfp4-2x-dgx-spark-sglang-tp2-roce","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":2,"vram_gb":128},"score":12.4000000000000003552713678800500929355621337890625,"score_basis":"speed-only","decode":{"tok_s":40.2999999999999971578290569595992565155029296875,"pct_of_generation_best":11.300000000000000710542735760100185871124267578125},"prefill":{"tok_s":2936.6999999999998181010596454143524169921875,"pct_of_generation_best":16.800000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-nvfp4-4x-dgx-spark-switchless-ring-tp4-dflash2","title":"GLM-5.3-Flash NVFP4 4x DGX Spark switchless ring TP4 DFlash2","tier":"community","url":"/r/glm-5-3-flash-nvfp4-4x-dgx-spark-switchless-ring-tp4-dflash2","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":4,"vram_gb":128},"score":12.300000000000000710542735760100185871124267578125,"score_basis":"speed-only","decode":{"tok_s":45,"pct_of_generation_best":12.5999999999999996447286321199499070644378662109375},"prefill":{"tok_s":1971.59999999999990905052982270717620849609375,"pct_of_generation_best":11.300000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-uncensored-nvfp4-bf16-ple-dgx-spark-vllm-mtp2","title":"Qwen3.8-Flash-Next Uncensored NVFP4 BF16 PLE DGX Spark vLLM MTP2","tier":"community","url":"/r/qwen3-8-flash-next-uncensored-nvfp4-bf16-ple-dgx-spark-vllm-mtp2","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":12.300000000000000710542735760100185871124267578125,"score_basis":"speed-only","decode":{"tok_s":44.2000000000000028421709430404007434844970703125,"pct_of_generation_best":12.4000000000000003552713678800500929355621337890625},"prefill":{"tok_s":2055.1999999999998181010596454143524169921875,"pct_of_generation_best":11.699999999999999289457264239899814128875732421875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-dgx-spark-gb10-sglang-mtp-213-262k","title":"Qwen3.8-Flash-Next NVFP4 DGX Spark GB10 SGLang MTP-213 262K","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-dgx-spark-gb10-sglang-mtp-213-262k","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":7.29999999999999982236431605997495353221893310546875,"score_basis":"speed-only","decode":{"tok_s":32.7000000000000028421709430404007434844970703125,"pct_of_generation_best":9.199999999999999289457264239899814128875732421875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-vq-q4-k-1x-dgx-spark-1m-ctx-jlfn","title":"GLM-5.3-Flash VQ Q4_K 1x DGX Spark 1M ctx (JLFN)","tier":"community","url":"/r/glm-5-3-flash-vq-q4-k-1x-dgx-spark-1m-ctx-jlfn","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":1,"vram_gb":128},"score":5.4000000000000003552713678800500929355621337890625,"score_basis":"speed-only","decode":{"tok_s":24.300000000000000710542735760100185871124267578125,"pct_of_generation_best":6.79999999999999982236431605997495353221893310546875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-1-flash-4x6000-kk-dspark-k7","title":"DeepSeek V4.1 Flash on 4x RTX PRO 6000 Blackwell \u2014 Karmic Kraken TP4 adaptive DSpark K7","tier":"community","url":"/r/ds4-1-flash-4x6000-kk-dspark-k7","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":4,"vram_gb":96},"score":0,"score_basis":"speed-only","decode":{"tok_s":0,"pct_of_generation_best":0},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-nvfp4-on-2x-dgx-spark-atlas-rc4-2-tp2-dflash2-1m","title":"GLM-5.3-Flash NVFP4 on 2x DGX Spark - Atlas RC4.2 TP2, DFlash2, 1M","tier":"community","url":"/r/glm-5-3-flash-nvfp4-on-2x-dgx-spark-atlas-rc4-2-tp2-dflash2-1m","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"dgx-spark","device_count":2,"vram_gb":128},"score":0,"score_basis":"speed-only","decode":{"tok_s":0,"pct_of_generation_best":0},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-flash-2x6000-kk-dspark-k5","title":"DeepSeek V4 Flash-0731 on 2x RTX PRO 6000 Blackwell \u2014 Karmic Kraken TP2 DSpark K5","tier":"community","url":"/r/ds4-flash-2x6000-kk-dspark-k5","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":2,"vram_gb":96},"score":0,"score_basis":"speed-only","decode":{"tok_s":0,"pct_of_generation_best":0},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-vision-2x6000-kk-dspark-k3","title":"DeepSeek V4 Flash Vision on 2x RTX PRO 6000 Blackwell \u2014 Karmic Kraken TP2 DSpark K3","tier":"community","url":"/r/ds4-vision-2x6000-kk-dspark-k3","generation":"blackwell","generation_label":"Blackwell","hardware":{"vendor":"nvidia","card":"RTX PRO 6000 Blackwell","device_count":2,"vram_gb":96},"score":0,"score_basis":"speed-only","decode":{"tok_s":0,"pct_of_generation_best":0},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-4x170hx-tp4-pp4-morrowmake-141","title":"GLM-5.3-Flash on 4x CMP 170HX \u2014 Morrowmake recipe 1.8.0, TP4 DFlash2","tier":"curated","url":"/r/glm-5-3-flash-4x170hx-tp4-pp4-morrowmake-141","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":89.400000000000005684341886080801486968994140625,"score_basis":"speed-only","decode":{"tok_s":264.3999999999999772626324556767940521240234375,"pct_of_generation_best":96.2000000000000028421709430404007434844970703125},"prefill":{"tok_s":6264,"pct_of_generation_best":62.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-uncensored-awq-w4a16-on-2x-cmp-170hx-8-slots","title":"Qwen3.8-Flash-Next-Uncensored AWQ W4A16 on 2x CMP 170HX \u2014 8 slots, PP2, MTP5 + int4 draft head, PLE-from-SSD (mannydelrio)","tier":"community","url":"/r/qwen3-8-flash-next-uncensored-awq-w4a16-on-2x-cmp-170hx-8-slots","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":83.7999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":240.400000000000005684341886080801486968994140625,"pct_of_generation_best":87.400000000000005684341886080801486968994140625},"prefill":{"tok_s":6964,"pct_of_generation_best":69.400000000000005684341886080801486968994140625},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"mimo-v2-6-flash-rl-309b-moe-nvfp4-4x-cmp-170hx-tp4-dflash-1m-ctx","title":"MiMo-V2.6-Flash-RL 309B MoE NVFP4 4x CMP 170HX TP4 DFlash 1M ctx (ElectroHobo)","tier":"community","url":"/r/mimo-v2-6-flash-rl-309b-moe-nvfp4-4x-cmp-170hx-tp4-dflash-1m-ctx","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":80,"score_basis":"speed-only","decode":{"tok_s":274.8999999999999772626324556767940521240234375,"pct_of_generation_best":100},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-2x-cmp-170hx-tp2-dflash2-host-ram-ple","title":"Qwen3.8-Flash-Next W4A16 2x CMP 170HX TP2 DFlash2 host-RAM PLE","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-2x-cmp-170hx-tp2-dflash2-host-ram-ple","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":73.7000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":253.19999999999998863131622783839702606201171875,"pct_of_generation_best":92.099999999999994315658113919198513031005859375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-fp8-4x-cmp-170hx-pp4-mtp-4-patched-vllm-g4bwy","title":"Qwen3.8-Flash-Next-FP8 4x CMP 170HX \u2014 PP4 MTP-4 patched vLLM (g4bwy)","tier":"community","url":"/r/qwen3-8-flash-next-fp8-4x-cmp-170hx-pp4-mtp-4-patched-vllm-g4bwy","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":73.2999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":183.19999999999998863131622783839702606201171875,"pct_of_generation_best":66.599999999999994315658113919198513031005859375},"prefill":{"tok_s":10038,"pct_of_generation_best":100},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-2x-cmp-170hx-pp2-mtp5-ple-alexbrandes","title":"Qwen3.8-Flash-Next W4A16 2x CMP 170HX PP2 MTP5 PLE (AlexBrandes)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-2x-cmp-170hx-pp2-mtp5-ple-alexbrandes","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":65.7999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":226.19999999999998863131622783839702606201171875,"pct_of_generation_best":82.2999999999999971578290569595992565155029296875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-merlin-on-4x-rtx-3090-tp2xpp2-ep-mtp3-f","title":"Qwen3.8-Flash-Next W4A16-Merlin on 4x RTX 3090 \u2014 TP2xPP2+EP, MTP3, FP8 KV, 925K pool (halt95)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-merlin-on-4x-rtx-3090-tp2xpp2-ep-mtp3-f","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":4,"vram_gb":24},"score":59.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":168.099999999999994315658113919198513031005859375,"pct_of_generation_best":61.10000000000000142108547152020037174224853515625},"prefill":{"tok_s":5360,"pct_of_generation_best":53.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-int4-mixed-170hx-x2-pp2-mtp3-bf16-kv-ple-nvme","title":"Qwen3.8-Flash-Next INT4-Mixed 170HX x2, PP2 MTP3 BF16-KV PLE-NVMe (bozocentral)","tier":"community","url":"/r/qwen3-8-flash-next-int4-mixed-170hx-x2-pp2-mtp3-bf16-kv-ple-nvme","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":57.89999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":199,"pct_of_generation_best":72.400000000000005684341886080801486968994140625},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-3bpw-fp8-ple-1x-cmp-170hx-dflash2-recoverssm","title":"Qwen3.8-Flash-Next 3bpw FP8-PLE 1x CMP 170HX DFlash2 RecoverSSM","tier":"community","url":"/r/qwen3-8-flash-next-3bpw-fp8-ple-1x-cmp-170hx-dflash2-recoverssm","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":1,"vram_gb":64},"score":57.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":197.80000000000001136868377216160297393798828125,"pct_of_generation_best":72},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-sm80-deepinfer-kernel-port-6x-a100-dspark-k5","title":"DeepSeek-V4.1-Flash SM80 DeepInfer kernel port 6x CMP 170HX DSpark K5","tier":"community","url":"/r/deepseek-v4-1-flash-sm80-deepinfer-kernel-port-6x-a100-dspark-k5","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":6,"vram_gb":64},"score":55.7999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":160.80000000000001136868377216160297393798828125,"pct_of_generation_best":58.5},"prefill":{"tok_s":4506.8000000000001818989403545856475830078125,"pct_of_generation_best":44.89999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-hybrid-exl3-tp4-4x-cmp-170hx-morrowmake","title":"DeepSeek-V4.1-Flash-1017 Hybrid EXL3 TP4 4x CMP 170HX Morrowmake","tier":"curated","url":"/r/deepseek-v4-1-flash-1017-hybrid-exl3-tp4-4x-cmp-170hx-morrowmake","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":54.7000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":162,"pct_of_generation_best":58.89999999999999857891452847979962825775146484375},"prefill":{"tok_s":3804,"pct_of_generation_best":37.89999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-exl3-2-9bpw-tc-moe-on-4x-cmp-170hx-tp4","title":"DeepSeek-V4.1-Flash-1017 EXL3 2.9bpw TC-MoE on 4x CMP 170HX TP4","tier":"community","url":"/r/deepseek-v4-1-flash-1017-exl3-2-9bpw-tc-moe-on-4x-cmp-170hx-tp4","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":51,"score_basis":"speed-only","decode":{"tok_s":165.599999999999994315658113919198513031005859375,"pct_of_generation_best":60.2000000000000028421709430404007434844970703125},"prefill":{"tok_s":1390.09999999999990905052982270717620849609375,"pct_of_generation_best":13.800000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-nvfp4-2x-cmp-170hx-tep2-gds-ssd-ple-mtp6","title":"Qwen3.8-Flash-Next NVFP4 2x CMP 170HX TEP2 GDS SSD PLE MTP6","tier":"community","url":"/r/qwen3-8-flash-next-nvfp4-2x-cmp-170hx-tep2-gds-ssd-ple-mtp6","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":50.10000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":158.19999999999998863131622783839702606201171875,"pct_of_generation_best":57.5},"prefill":{"tok_s":2059.90000000000009094947017729282379150390625,"pct_of_generation_best":20.5},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-exl3-5-05bpw-2x-cmp-170hx-tp2-mtp3-morrowmake","title":"Qwen3.8-Flash-Next EXL3 5.05bpw 2x CMP 170HX TP2 MTP3 (Morrowmake 1.0.0)","tier":"community","url":"/r/qwen3-8-flash-next-exl3-5-05bpw-2x-cmp-170hx-tp2-mtp3-morrowmake","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":49.7000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":152.80000000000001136868377216160297393798828125,"pct_of_generation_best":55.60000000000000142108547152020037174224853515625},"prefill":{"tok_s":2651,"pct_of_generation_best":26.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-uncensored-awq-w4a16-on-2x-cmp-170hx-pp2-mtp3","title":"Qwen3.8-Flash-Next-Uncensored AWQ W4A16 on 2x CMP 170HX \u2014 PP2, MTP3, PLE-from-SSD (mannydelrio)","tier":"community","url":"/r/qwen3-8-flash-next-uncensored-awq-w4a16-on-2x-cmp-170hx-pp2-mtp3","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":47.10000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":109,"pct_of_generation_best":39.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":7700,"pct_of_generation_best":76.7000000000000028421709430404007434844970703125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-exl3-3bpw-tc-moe-on-4x-cmp-170hx-tp4","title":"DeepSeek-V4.1-Flash EXL3 3bpw TC-MoE on 4x CMP 170HX TP4 (original checkpoint)","tier":"community","url":"/r/deepseek-v4-1-flash-exl3-3bpw-tc-moe-on-4x-cmp-170hx-tp4","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":46.39999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":149.19999999999998863131622783839702606201171875,"pct_of_generation_best":54.2999999999999971578290569595992565155029296875},"prefill":{"tok_s":1486,"pct_of_generation_best":14.800000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-engram-nvme-tp4-pp4-4x-cmp-170hx","title":"DeepSeek-V4.1-Flash-1017 Engram-NVMe TP4/PP4 4x CMP 170HX","tier":"community","url":"/r/deepseek-v4-1-flash-1017-engram-nvme-tp4-pp4-4x-cmp-170hx","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":43.7999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":150.599999999999994315658113919198513031005859375,"pct_of_generation_best":54.7999999999999971578290569595992565155029296875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"ds4-flash-3x170hx-pp3-dspark-k5","title":"DeepSeek V4 Flash-0731 abliterated on 3x CMP 170HX \u2014 vLLM patched PP3 DSpark K5","tier":"curated","url":"/r/ds4-flash-3x170hx-pp3-dspark-k5","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":3,"vram_gb":64},"score":42.5,"score_basis":"speed-only","decode":{"tok_s":129.900000000000005684341886080801486968994140625,"pct_of_generation_best":47.2999999999999971578290569595992565155029296875},"prefill":{"tok_s":2350,"pct_of_generation_best":23.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-awq-int4-dspark-on-rtx-3090s-sglang-tp1-tp2-tp4","title":"Qwen3.8-27B AWQ-INT4 + DSpark on RTX 3090s \u2014 SGLang TP1/TP2/TP4","tier":"community","url":"/r/qwen3-8-27b-awq-int4-dspark-on-rtx-3090s-sglang-tp1-tp2-tp4","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":4,"vram_gb":24},"score":39.7000000000000028421709430404007434844970703125,"score_basis":"speed-only","decode":{"tok_s":127.099999999999994315658113919198513031005859375,"pct_of_generation_best":46.2000000000000028421709430404007434844970703125},"prefill":{"tok_s":1372.700000000000045474735088646411895751953125,"pct_of_generation_best":13.699999999999999289457264239899814128875732421875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-int4ple-on-2x-modded-cmp-170hx-pp2-mtp3","title":"Qwen3.8-Flash-Next W4A16 INT4PLE on 2x modded CMP 170HX \u2014 PP2 MTP3 (DUFRelic)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-int4ple-on-2x-modded-cmp-170hx-pp2-mtp3","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":38.7999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":133.400000000000005684341886080801486968994140625,"pct_of_generation_best":48.5},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-high-effort-on-2x-cmp170hx-exl3-tp2-four-slots","title":"GLM-5.3-Flash high effort on 2x CMP170HX EXL3 TP2 four slots","tier":"community","url":"/r/glm-5-3-flash-high-effort-on-2x-cmp170hx-exl3-tp2-four-slots","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":38.60000000000000142108547152020037174224853515625,"score_basis":"speed+quality","decode":{"tok_s":43.2000000000000028421709430404007434844970703125,"pct_of_generation_best":15.699999999999999289457264239899814128875732421875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":43.162181670367999686277471482753753662109375,"pct_of_generation_best":100}},{"slug":"glm-5-3-flash-exl3-3-05bpw-on-2x-cmp-170hx-tp2-dflash2-k7-hybrid","title":"GLM-5.3-Flash EXL3 3.05bpw on 2x CMP 170HX TP2 DFlash2 K7 hybrid drafting 393K (alexvnesta)","tier":"community","url":"/r/glm-5-3-flash-exl3-3-05bpw-on-2x-cmp-170hx-tp2-dflash2-k7-hybrid","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":38.5,"score_basis":"speed-only","decode":{"tok_s":123.599999999999994315658113919198513031005859375,"pct_of_generation_best":45},"prefill":{"tok_s":1291.5,"pct_of_generation_best":12.9000000000000003552713678800500929355621337890625},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-low-effort-on-2x-cmp170hx-exl3-tp2-four-slots","title":"GLM-5.3-Flash low effort on 2x CMP170HX EXL3 TP2 four slots","tier":"community","url":"/r/glm-5-3-flash-low-effort-on-2x-cmp170hx-exl3-tp2-four-slots","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":37.2999999999999971578290569595992565155029296875,"score_basis":"speed+quality","decode":{"tok_s":41.7000000000000028421709430404007434844970703125,"pct_of_generation_best":15.199999999999999289457264239899814128875732421875},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":41.7217847448689980183189618401229381561279296875,"pct_of_generation_best":96.7000000000000028421709430404007434844970703125}},{"slug":"glm-5-3-flash-max-effort-on-2x-cmp170hx-exl3-tp2-four-slots","title":"GLM-5.3-Flash max effort on 2x CMP170HX EXL3 TP2 four slots","tier":"community","url":"/r/glm-5-3-flash-max-effort-on-2x-cmp170hx-exl3-tp2-four-slots","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":37.10000000000000142108547152020037174224853515625,"score_basis":"speed+quality","decode":{"tok_s":41.5,"pct_of_generation_best":15.0999999999999996447286321199499070644378662109375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":41.47331427713900353637654916383326053619384765625,"pct_of_generation_best":96.099999999999994315658113919198513031005859375}},{"slug":"qwen3-8-flash-next-w4a16-fp8ple-on-2x-rtx-3090-tp2-expert-offloa","title":"Qwen3.8-Flash-Next W4A16-FP8PLE on 2x RTX 3090 \u2014 TP2, expert-offload, MTP3 (DominikBucko)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-fp8ple-on-2x-rtx-3090-tp2-expert-offloa","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":2,"vram_gb":24},"score":36.89999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":106.5,"pct_of_generation_best":38.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":2947,"pct_of_generation_best":29.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-3bpw-mtp-on-1x-cmp-170hx-ssd-ple-aio-iiiusi0n","title":"Qwen3.8-Flash-Next 3bpw MTP on 1x CMP 170HX SSD PLE AIO (iIIusi0n)","tier":"community","url":"/r/qwen3-8-flash-next-3bpw-mtp-on-1x-cmp-170hx-ssd-ple-aio-iiiusi0n","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":1,"vram_gb":64},"score":36.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":111,"pct_of_generation_best":40.39999999999999857891452847979962825775146484375},"prefill":{"tok_s":2132,"pct_of_generation_best":21.199999999999999289457264239899814128875732421875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-fidelity-first-4x-cmp-170hx-alanoo81","title":"DeepSeek-V4.1-Flash-1017 Fidelity-First 4x CMP 170HX alanoo81","tier":"community","url":"/r/deepseek-v4-1-flash-1017-fidelity-first-4x-cmp-170hx-alanoo81","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":35.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":105,"pct_of_generation_best":38.2000000000000028421709430404007434844970703125},"prefill":{"tok_s":2514,"pct_of_generation_best":25},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-2x-cmp-170hx-tp2-ep-mtp4-fp8-ple-92gb-h","title":"Qwen3.8-Flash-Next W4A16 2x CMP 170HX TP2+EP MTP4 FP8 PLE 92GB Host","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-2x-cmp-170hx-tp2-ep-mtp4-fp8-ple-92gb-h","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":35.10000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":101.2999999999999971578290569595992565155029296875,"pct_of_generation_best":36.7999999999999971578290569595992565155029296875},"prefill":{"tok_s":2839,"pct_of_generation_best":28.300000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-fp8ple-2x-cmp-170hx-pp2-mtp4-1m","title":"Qwen3.8-Flash-Next W4A16-FP8PLE 2x CMP 170HX PP2 MTP4 1M","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-fp8ple-2x-cmp-170hx-pp2-mtp4-1m","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":31.800000000000000710542735760100185871124267578125,"score_basis":"speed-only","decode":{"tok_s":69.7999999999999971578290569595992565155029296875,"pct_of_generation_best":25.39999999999999857891452847979962825775146484375},"prefill":{"tok_s":5750,"pct_of_generation_best":57.2999999999999971578290569595992565155029296875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-3x-cmp-170hx-pp3-nvme-ple-fp8-kv-1m-cle","title":"Qwen3.8-Flash-Next W4A16 3x CMP 170HX PP3 NVMe PLE fp8 KV 1M (clee)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-3x-cmp-170hx-pp3-nvme-ple-fp8-kv-1m-cle","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":3,"vram_gb":64},"score":28.39999999999999857891452847979962825775146484375,"score_basis":"speed-only","decode":{"tok_s":73.2000000000000028421709430404007434844970703125,"pct_of_generation_best":26.60000000000000142108547152020037174224853515625},"prefill":{"tok_s":3563,"pct_of_generation_best":35.5},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exl3-3-05bpw-on-2x-cmp-170hx-dflash2-k7-q8-kv-384k","title":"GLM-5.3-Flash EXL3 3.05bpw on 2x CMP 170HX \u2014 DFlash2 K7 Q8 KV 384K (Flun)","tier":"community","url":"/r/glm-5-3-flash-exl3-3-05bpw-on-2x-cmp-170hx-dflash2-k7-q8-kv-384k","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":2,"vram_gb":64},"score":26.199999999999999289457264239899814128875732421875,"score_basis":"speed-only","decode":{"tok_s":90,"pct_of_generation_best":32.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"mimo-v2-6-flash-rl-omni-on-4x-cmp-170hx-pp4-official-fp8-three-t","title":"MiMo-V2.6-Flash-RL (omni) on 4x CMP 170HX \u2014 PP4, official fp8, three-tier KV, v3.3 (kaka86mm)","tier":"community","url":"/r/mimo-v2-6-flash-rl-omni-on-4x-cmp-170hx-pp4-official-fp8-three-t","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":24.699999999999999289457264239899814128875732421875,"score_basis":"speed-only","decode":{"tok_s":60,"pct_of_generation_best":21.800000000000000710542735760100185871124267578125},"prefill":{"tok_s":3650,"pct_of_generation_best":36.39999999999999857891452847979962825775146484375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-int8-mtp3-prefix-cache-on-1x-cmp-170hx-hynix","title":"Qwen3.8-27B INT8 MTP3 prefix-cache on 1x CMP 170HX Hynix","tier":"community","url":"/r/qwen3-8-27b-int8-mtp3-prefix-cache-on-1x-cmp-170hx-hynix","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":1,"vram_gb":64},"score":20,"score_basis":"speed-only","decode":{"tok_s":68.7000000000000028421709430404007434844970703125,"pct_of_generation_best":25},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-flash-next-w4a16-fp8ple-on-1x-rtx-3090-cpu-cold-experts","title":"Qwen3.8-Flash-Next W4A16-FP8PLE on 1x RTX 3090 \u2014 CPU cold experts, MTP3, 128K (DominikBucko)","tier":"community","url":"/r/qwen3-8-flash-next-w4a16-fp8ple-on-1x-rtx-3090-cpu-cold-experts","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":1,"vram_gb":24},"score":18.5,"score_basis":"speed-only","decode":{"tok_s":48.89999999999999857891452847979962825775146484375,"pct_of_generation_best":17.800000000000000710542735760100185871124267578125},"prefill":{"tok_s":2123.5,"pct_of_generation_best":21.199999999999999289457264239899814128875732421875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-764b-on-4x-cmp-170hx-pp4-shadow-kv-exl3-2bpw","title":"DeepSeek-V4.1-Flash (764B) on 4x CMP 170HX \u2014 PP4, shadow KV, EXL3 2bpw, DSpark K5, 512k ctx, native vision (kaka86mm)","tier":"community","url":"/r/deepseek-v4-1-flash-764b-on-4x-cmp-170hx-pp4-shadow-kv-exl3-2bpw","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"170hx","device_count":4,"vram_gb":64},"score":16.699999999999999289457264239899814128875732421875,"score_basis":"speed-only","decode":{"tok_s":57.5,"pct_of_generation_best":20.89999999999999857891452847979962825775146484375},"prefill":{"tok_s":0,"pct_of_generation_best":0},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exl3-3-05bpw-on-2x-rtx-3090-vllm-exllamav3-cpu-exp","title":"GLM-5.3-Flash EXL3 3.05bpw on 2x RTX 3090 \u2014 vLLM + ExLlamaV3 CPU expert offload, 256K (DominikBucko)","tier":"community","url":"/r/glm-5-3-flash-exl3-3-05bpw-on-2x-rtx-3090-vllm-exllamav3-cpu-exp","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":2,"vram_gb":24},"score":16.60000000000000142108547152020037174224853515625,"score_basis":"speed-only","decode":{"tok_s":49.39999999999999857891452847979962825775146484375,"pct_of_generation_best":18},"prefill":{"tok_s":1130.5,"pct_of_generation_best":11.300000000000000710542735760100185871124267578125},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"glm-5-3-flash-exl3-3-05bpw-on-1x-rtx-3090-ram-nvme-expert-tierin","title":"GLM-5.3-Flash EXL3 3.05bpw on 1x RTX 3090 \u2014 RAM/NVMe expert tiering (near-exact)","tier":"community","url":"/r/glm-5-3-flash-exl3-3-05bpw-on-1x-rtx-3090-ram-nvme-expert-tierin","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":1,"vram_gb":24},"score":9,"score_basis":"speed-only","decode":{"tok_s":24.10000000000000142108547152020037174224853515625,"pct_of_generation_best":8.800000000000000710542735760100185871124267578125},"prefill":{"tok_s":1015,"pct_of_generation_best":10.0999999999999996447286321199499070644378662109375},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"deepseek-v4-1-flash-1017-dsx-2-6-bit-gptq-2x-rtx-3090-engram","title":"DeepSeek-V4.1-Flash-1017 dsx 2.6-bit GPTQ 2x RTX 3090 (Engram)","tier":"community","url":"/r/deepseek-v4-1-flash-1017-dsx-2-6-bit-gptq-2x-rtx-3090-engram","generation":"ampere","generation_label":"Ampere","hardware":{"vendor":"nvidia","card":"rtx-3090","device_count":2,"vram_gb":24},"score":7.9000000000000003552713678800500929355621337890625,"score_basis":"speed-only","decode":{"tok_s":26,"pct_of_generation_best":9.5},"prefill":{"tok_s":170,"pct_of_generation_best":1.6999999999999999555910790149937383830547332763671875},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-paro-mxfp6-2x-radeon-ai-pro-r9700-vllm-radiance-tp2","title":"Qwen3.8-27B PARO-MXFP6 2x Radeon AI PRO R9700 \u2014 vLLM radiance TP2","tier":"community","url":"/r/qwen3-8-27b-paro-mxfp6-2x-radeon-ai-pro-r9700-vllm-radiance-tp2","generation":"amd","generation_label":"AMD","hardware":{"vendor":"amd","card":"radeon-ai-pro-r9700","device_count":2,"vram_gb":32},"score":100,"score_basis":"speed-only","decode":{"tok_s":138,"pct_of_generation_best":100},"prefill":{"tok_s":3348,"pct_of_generation_best":100},"quality":{"score":null,"pct_of_generation_best":0}},{"slug":"qwen3-8-27b-mxfp4-dflash2-1x-radeon-ai-pro-r9700-wsl2-radiance","title":"Qwen3.8-27B MXFP4 DFlash2 1x Radeon AI PRO R9700 WSL2 radiance","tier":"community","url":"/r/qwen3-8-27b-mxfp4-dflash2-1x-radeon-ai-pro-r9700-wsl2-radiance","generation":"amd","generation_label":"AMD","hardware":{"vendor":"amd","card":"radeon-ai-pro-r9700","device_count":1,"vram_gb":32},"score":92.2999999999999971578290569595992565155029296875,"score_basis":"speed-only","decode":{"tok_s":130.69999999999998863131622783839702606201171875,"pct_of_generation_best":94.7000000000000028421709430404007434844970703125},"prefill":{"tok_s":2762,"pct_of_generation_best":82.5},"quality":{"score":null,"pct_of_generation_best":0}}]}