Drilldown: Recipe

DeepSeek V4.1 Flash heretic on 4x 8 GB (1) · DeepSeek V4.1 Flash on 6x 8 GB (TP6) (1) · DeepSeek V4.1 Flash on 8x 8 GB (Deepinfer-SM80 TP4 x PP2) (1) · GLM 5.3 Flash on 8x 8 GB (promisezackr PP8 NVFP4 DFlash2) (1) · Qwen3.8 Flash Next on 2x 8 GB (Morrowmake EXL3) (1) · Qwen3.8 Flash Next on 2x 8 GB (vektorprime PP2) (1) · Qwen3.8 Flash Next on 4x 8 GB (1)

DeepSeek-V4.1-Flash-1017 (2) · DeepSeek-V4.1-Flash-1017 heretic (pqhaz abliteration) (1) · GLM-5.3-Flash (1) · Qwen3.8-Flash-Next (3)

AWQ INT4: cyankiwi/Qwen3.8-Flash-Next-AWQ-INT4; 95 GiB BF16 PLE table in host RAM (FP8 table optional) (1) · AWQ W4A16: wtdcode/Qwen3.8-Flash-Next-AWQ-W4A16 (1) · EXL3 5.05 bpw; PLE (n-gram) table read from SSD (1) · Hybrid: routed experts EXL3 2.9 bpw (LibertAIDAI), everything else official FP8/BF16 with the 38 heretic attn.wo_b tensors (1) · LibertAIDAI/GLM-5.3-Flash-NVFP4 (ModelOpt NVFP4, routed experts only, 181 GiB), run through Marlin W4A16 kernels; GLM-5.3-Flash DFlash2 drafter (5 sliding-window layers) (1) · Official checkpoint (FP8 dense, FP4 experts); software FP8 on sm_80 (1) · Official checkpoint: dense MXFP8 through an SM80 Marlin W8A16 path, routed experts MXFP4 through W4A16; KV FP8 (1)

Custom SM80 vLLM (patched from public commit 79a7108d9) (1) · Morrowmake/vllm-cmp170hx (EXL3 path) (1) · Morrowmake/vllm-cmp170hx (vLLM fork, container image) (1) · vLLM nightly + PR #53899 (a5530b9) + the repo's patch scripts (1) · vLLM PR #56201 head (79a7108d) plus the two Deepinfer-SM80 patches and 30 drop-in modules (1) · vLLM: official vllm/vllm-openai:glm53-flash image plus promisezackr's 24 patches (SM8x base from 344303947/vllm170hx, pipeline parallel for GLM-5.3-Flash, DFlash2 under PP8) (1) · wtdcode/vllm-backport (image lazymio/vllm-backport:latest-sm80) (1)

0.1.dev1+ga5530b90c.d20260829 (1) · Deepinfer-SM80, vLLM v0.28.1rc0 base (1) · latest-sm80 tag as of 2026-10-16 (1) · recipe 1.0.0 (commit d4d1f525), engine a0c57b52 (1) · reports 0.0.0; patches/public-base-to-runtime.patch (109 files) (1) · tag qwen38-recipe-1.0.0, engine 18305948ac (1) · v0.1.dev20051+g487ecf187 (the image's vLLM); repo commit 90ec72e (1)

8 GB (7)

2 (2) · 4 (2) · 6 (1) · 8 (2)

PP2 with PLE CPU offload (1) · PP8 (VLLM_PP_LAYER_PARTITION=7,5,7,5,7,5,7,2) with DFlash2, 7 draft tokens; PCIe Gen2 x4, no P2P; host 2x Xeon E5-2690 v4, 96 GB DDR4 (1) · TP2 with expert parallel (1) · TP4 (1) · TP4 with expert parallel, MTP 3 (1) · TP4 x PP2, partition 20,20 (1) · TP6 (virtual-head padding), PP1 (1)

1,000,000 (YaRN x4 from 262,144) (1) · 1,048,576 (2) · 1,048,576 (KV pool 6.62M tokens at utilization 0.90, about 6.3 concurrent 1M requests) (1) · 262,144 (3)

70 - 120 (3) · 120 - 160 (4)

None (3) · 2,101.4 (1) · 2,453 (1) · 2,651 (1) · 3,463 (1)

64 GB (1) · 64 GB, --gpu-memory-utilization 0.90 (1) · 64 GB, --gpu-memory-utilization 0.95 (1) · 64 GiB, --gpu-memory-utilization 0.92 (miner co-resident) (1) · 64 GiB, --gpu-memory-utilization 0.95 (2) · 64 GiB; weights PP0 39.48 GiB, PP1 45.53 GiB; --gpu-memory-utilization 0.9 (1)

None · 180 W · 200 W · 100 W · 140 W · 140 W (200 W observed-draw abort guard) · 150 W per CMP (PRO 400 W) · 170 W (author's 10-card run) · 180 W (author's cards); about 170 W with a modest memory OC in his Discord numbers · 180 W (nvidia-smi -pl 180) · 180 W (one card unstable at 250 W) · Other

None (7)

None (6) · cmpunlocker VRAM unlock (64 GB per card) (1)

Reported · Proven · Other · None

darkmaniac7 (1) · HalfVulpes (1) · matstatman (1) · promisezackr (1) · vektor_prime (1) · voidgiraffe (1) · wgkbfcp (1)

August 2026 (2) · October 2026 (5)

https://discord.com/channels/1518634379604136178/1528826330454429706/1560602001908170753 (1) · https://github.com/HalfVulpes/Deepinfer-SM80 (1) · https://github.com/Morrowmake/dsv41-flash-cmp170hx-recipe (1) · https://github.com/Morrowmake/qwen38-flash-next-cmp170hx-recipe (1) · https://github.com/promisezackr/glm53-flash-170hx-pp8 (1) · https://github.com/vektorprime/qwen38-flash-next-pp2 (1) · https://github.com/wgkbfcp/deepseek-v4.1-flash-1017-cmp170hx-recipe (1)