Drilldown: Recipe

DeepSeek V4.1 Flash on 2x 8 GB (Project Maya) (1) · GLM 5.3 Flash on 2x 8 GB (Project Maya) (1) · Qwen3.8 27B on 1x 8 GB (W4A16 vLLM) (1) · Qwen3.8 Flash Next on 2x 8 GB (ElectroHobo W4A16 TP2) (1) · Qwen3.8 Flash Next on 2x 8 GB (Plean PP2) (1) · Qwen3.8 Flash Next on 4x 8 GB (1)

DeepSeek-V4.1-Flash-1017 (748B in the Maya README) (1) · GLM-5.3-Flash (321B in the Maya README) (1) · Qwen3.8-27B (dense) (1) · Qwen3.8-Flash-Next (2) · Qwen3.8-Flash-Next (125B MoE per the recipe) (1)

AWQ W4A16: wtdcode/Qwen3.8-Flash-Next-AWQ-W4A16 (1) · Maya GGUF (Maya-S 340 GB or Maya-XS 277 GB; Engram tables stay on SSD) (1) · Maya-M GGUF (116 GB), every expert in VRAM; Maya-S (96.5 GB) and Maya-S24 (94.7 GB) also reported (1) · Mixed INT4 AutoRound (Minachist/Qwen3.8-Flash-Next-INT4-Mixed-AutoRound, about 65 GB) with the full BF16 PLE table on NVMe (1) · W4A16 (INT4 weights, Marlin kernels) (1) · W4A16: halt95/Qwen3.8-Flash-Next-W4A16-Merlin (AutoRound INT4 g128 experts, INT8 GDN projections, FP8 n-gram table) (1)

Morrowmake/vllm-cmp170hx image (Qwen EXL3 release) with patch files mounted read-only (1) · Project Maya (Strata and llama.cpp lineage) (1) · Project Maya (Strata and llama.cpp lineage, EXL3 loading, DFlash2 drafter) (1) · vLLM (1) · vLLM 0.29.1rc1 nightly plus the checkpoint author's patches (1) · wtdcode/vllm-backport (image lazymio/vllm-backport:latest-sm80) (1)

None (3) · latest-sm80 tag as of 2026-10-16 (1) · recipe 0.3.1; engine 18305948ac (1) · v1.0.6 (this row) (1)

8 GB · Other · None

1 (1) · 2 (4) · 4 (1)

engine-managed expert tiering across VRAM, host RAM and NVMe (1) · engine-managed split with expert tiering across VRAM, RAM and NVMe (1) · PP2, MTP 3, no P2P (1) · single card (1) · TP2 with whole-expert expert parallel (1) · TP4 with expert parallel, MTP 3 (1)

None (2) · 1,000,000 (YaRN x4 from 262,144) (1) · 262,144 (1) · 262,144 (256K in the post) (1) · 262,144 native (YaRN to about 500k in the recipe's long variant) (1)

0 - 400 (6)

None (2) · 514 (1) · 2,101.4 (1) · 4,696 (1) · 5,800 (1)

64 GB · 64 GiB · 64 GiB exposed, --gpu-memory-utilization 0.95 · 64 GiB, --gpu-memory-utilization 0.95 · 64 GB, --gpu-memory-utilization 0.94, TB_TORCH_MEM_CAP_GIB=61.5 · 64 GB, --gpu-memory-utilization 0.95 · 64 GB; weights 35.9 GB per rank (ranks 1-6), 37.6 GB rank 0, 46.3 GB rank 7 · 64 GiB (64,911 MiB usable), GPU_MEMORY_UTILIZATION 0.98 · 64 GiB, --gpu-memory-utilization 0.90 · 64 GiB, --gpu-memory-utilization 0.92 (miner co-resident) · 64 GiB, --gpu-memory-utilization 0.95 (fixed) · 64 GiB, --gpu-memory-utilization 0.96 · 64 GiB, --gpu-memory-utilization 0.97 · 64 GiB, gpu-memory-utilization 0.92 · 64 GiB, GPU_MEMORY_UTILIZATION 0.95 · 64 GiB, GPU_UTIL 0.95 (fixed) · 64 GiB, GPU_UTIL 0.96 at PP8 · 64 GiB; weights 45.5 GiB plus 10 GiB KV per card; GPU_UTIL 0.94 (fixed) · 64 GiB; weights PP0 39.48 GiB, PP1 45.53 GiB; --gpu-memory-utilization 0.9 · about 58 GiB of 64 GiB (116.6 GiB resident across both) · about 62.5 GB used of 64 GB, --gpu-memory-utilization 0.97 · about 62.9 / 62.3 GiB idle, peaks about 64.4 / 63.0 GiB with vision · gpu_split 58 GB per CMP, 70 GB on the PRO · Other · None

None (5) · 180 W (1)

None (5) · 580 or newer (measured on 610.57.04); open modules with HMM for the table-on-disk mode (1)

None (5) · cmpunlocker (64 GiB, 74 SMs) with a P2P-capable build (1)

Proven (1) · Reported (5)

kellertreppe123 (1) · matstatman (1) · nathang6866 (1) · pcsv. (1) · plean. (1) · ZackO2o (1)

October 5, 2026 (1) · October 7, 2026 (1) · October 16, 2026 (1) · October 23, 2026 (1) · October 25, 2026 (1)

https://discord.com/channels/1518634379604136178/1528826330454429706/1560602001908170753 (1) · https://github.com/ElectroHobo/qwen38-flash-next-w4a16-2x-cmp170hx-recipe (1) · https://github.com/mw00/project-maya#how-fast-is-it (2) · https://www.bozocentral.com/2x-cmp170hx-qwen3-flash-next (1) · https://www.reddit.com/r/LocalLLM/comments/1w2f03e/qwen3827b_w4a16_on_one_64_gb_cmp_170hx_16070/ (1)