Drilldown: Recipe - CMP Wiki

Jump to content

Choose a table:

Recipe > page

: GLM 5.3 Flash on 2x 8 GB (ExLlamaV3 TP2)

or Qwen3.8 Flash Next on 3x 8 GB (clee PP3)

& cards: 8 GB

Use the filters below to narrow your results.

GLM-5.3-Flash (320B MoE, 18B active) (1) · Qwen3.8-Flash-Next (1)

EXL3 3.05 bpw: turboderp/GLM-5.3-Flash-exl3; DFlash2 drafter converted to EXL3 6 bpw (1) · W4A16 AutoRound: Intel/Qwen3.8-Flash-Next-W4A16-AutoRound; 95.4 GiB BF16 PLE table streamed from NVMe; KV fp8 e4m3 (software-decoded on SM80) (1)

ExLlamaV3 (patched, pinned) (1) · vLLM nightly 7d0b4e57 plus the author's pure-Python fork clee/vllm-sm80 (branch q38fn/stack) (1)

q38fn/stack on nightly-7d0b4e57 (1) · upstream dev 0f685aa9fd plus 9 patches; repo db01c256 (1)

PP3, no P2P (works at PP2 on two cards) (1) · TP2 over NCCL (GPU P2P), DFlash2 K7 plus hybrid drafting (1)

1,048,576 (YaRN x4) (1) · 392,960 per request; 655,360-token Q8 KV pool shared by two requests (1)

74.4 (1) · 107.67 (1)

1,533 (1) · 4,831 (1)

64 GiB, gpu-memory-utilization 0.92 (1) · about 62.9 / 62.3 GiB idle, peaks about 64.4 / 63.0 GiB with vision (1)

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

None (1) · 610.57.04 (1)

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

Proven (1) · Reported (1)

c133 (1) · nestaa51 (1)

October 10, 2026 (1) · October 13, 2026 (1)

https://github.com/alexvnesta/glm53-flash-cmp170hx-tp (1) · https://github.com/clee/qwen3.8-flash-next-3x170hx (1)

Showing below up to 2 results in range #1 to #2.

G

GLM 5.3 Flash on 2x 8 GB (ExLlamaV3 TP2)

Q

Qwen3.8 Flash Next on 3x 8 GB (clee PP3)

Retrieved from "https://cmp.wiki/wiki/Special:Drilldown/Recipe"