Drilldown: Recipe

DeepSeek V4.1 Flash heretic on 4x 8 GB · DeepSeek V4.1 Flash on 2x 8 GB (Project Maya) · DeepSeek V4.1 Flash on 4x 8 GB (alanoo fidelity build) · DeepSeek V4.1 Flash on 4x 8 GB (CurtisAccelerate TC-MoE) · DeepSeek V4.1 Flash on 4x 8 GB (GNmac) · DeepSeek V4.1 Flash on 4x 8 GB (Morrowmake) · DeepSeek V4.1 Flash on 4x 8 GB plus RTX PRO 6000 (ExLlamaV3 TP5) · DeepSeek V4.1 Flash on 6x 8 GB (Deepinfer-SM80 TP2 x PP3) · DeepSeek V4.1 Flash on 6x 8 GB (TP6) · DeepSeek V4.1 Flash on 8x 8 GB (Deepinfer-SM80 TP4 x PP2) · DeepSeek V4.1 Flash on 8x 8 GB (Zanooda PP8) · GLM 5.3 Flash on 2x 8 GB (ExLlamaV3 layer split) · GLM 5.3 Flash on 2x 8 GB (ExLlamaV3 TP2) · GLM 5.3 Flash on 2x 8 GB (Project Maya) · GLM 5.3 Flash on 3x 8 GB · GLM 5.3 Flash on 4x 8 GB · GLM 5.3 on 8x 8 GB (Int4-Int8Mix PP8) · GLM 5.3 on 8x 8 GB (vllm-170hx-glm5) · MiMo V2.6 Flash on 4x 8 GB · Qwen3.8 27B on 1x 8 GB (W4A16 vLLM) · Qwen3.8 Flash Next on 1x 8 GB (ElectroHobo 3 bpw) · Qwen3.8 Flash Next on 1x 8 GB (iIIusi0n SSD PLE) · Qwen3.8 Flash Next on 2x 8 GB (DUFRelic PP2) · Qwen3.8 Flash Next on 2x 8 GB (ElectroHobo W4A16 TP2) · Qwen3.8 Flash Next on 2x 8 GB (mannydelrio PP2) · Qwen3.8 Flash Next on 2x 8 GB (Morrowmake EXL3) · Qwen3.8 Flash Next on 2x 8 GB (Plean PP2) · Qwen3.8 Flash Next on 2x 8 GB (vektorprime PP2) · Qwen3.8 Flash Next on 3x 8 GB (clee PP3) · Qwen3.8 Flash Next on 4x 8 GB · Other · None

DeepSeek-V4.1-Flash-1017 (748B in the Maya README) (1) · GLM-5.3-Flash (320B MoE, 18B active) (1) · Qwen3.8-27B (dense) (1)

EXL3 3.05 bpw: turboderp/GLM-5.3-Flash-exl3; DFlash2 drafter converted to EXL3 6 bpw (1) · Maya GGUF (Maya-S 340 GB or Maya-XS 277 GB; Engram tables stay on SSD) (1) · W4A16 (INT4 weights, Marlin kernels) (1)

ExLlamaV3 (patched, pinned) (1) · Project Maya (Strata and llama.cpp lineage) (1) · vLLM (1)

None (2) · upstream dev 0f685aa9fd plus 9 patches; repo db01c256 (1)

8 GB (3)

1 (1) · 2 (2)

engine-managed expert tiering across VRAM, host RAM and NVMe (1) · single card (1) · TP2 over NCCL (GPU P2P), DFlash2 K7 plus hybrid drafting (1)

None (1) · 262,144 native (YaRN to about 500k in the recipe's long variant) (1) · 392,960 per request; 655,360-token Q8 KV pool shared by two requests (1)

63 (1) · 107.67 (1) · 170 (1)

None (2) · 1,533 (1)

64 GB (2) · about 62.9 / 62.3 GiB idle, peaks about 64.4 / 63.0 GiB with vision (1)

None (2) · 180 W (author's cards); about 170 W with a modest memory OC in his Discord numbers (1)

None (2) · 610.57.04 (1)

None (2) · cmpunlocker (64 GiB); P2P-capable driver for the NCCL backend (1)

Reported (3)

kellertreppe123 (1) · nestaa51 (1) · pcsv. (1)

October 5, 2026 (1) · October 13, 2026 (1) · October 25, 2026 (1)

https://github.com/alexvnesta/glm53-flash-cmp170hx-tp (1) · https://github.com/mw00/project-maya#how-fast-is-it (1) · https://www.reddit.com/r/LocalLLM/comments/1w2f03e/qwen3827b_w4a16_on_one_64_gb_cmp_170hx_16070/ (1)