Page values for "DeepSeek V4.1 Flash on 4x 8 GB (Morrowmake)"
From CMP Wiki
More actions
"Recipe" values
1 row is stored for this page| Field | Field type | Allowed values | Value |
|---|---|---|---|
| page | Page | DeepSeek V4.1 Flash on 4x 8 GB (Morrowmake) | |
| model | String | DeepSeek-V4.1-Flash-1017 | |
| quant | String | Hybrid: routed experts EXL3 2.9 bpw (LibertAIDAI), everything else official FP8/BF16; KV BF16 | |
| engine | String | Morrowmake/vllm-cmp170hx (vLLM fork, container image) | |
| engine_version | String | recipe 1.0.0, engine a0c57b5237 | |
| cards | String | 8 GB · 10 GB · both | 8 GB |
| card_count | Integer | 4 | |
| topology | String | TP4 with expert parallel (96 whole experts per card) | |
| context | String | 262,144 | |
| decode_tps | Float | 287.0 | |
| prefill_tps | Float | 3,818.0 | |
| vram_per_card | String | 64 GiB exposed, --gpu-memory-utilization 0.95 | |
| power_cap | String | 180 W | |
| driver | String | NVIDIA 580 or newer; P2P needs the Morrowmake/cmpunlocker build | |
| unlock_build | String | cmpunlocker (64 GiB, 74 SMs); Morrowmake/cmpunlocker for P2P | |
| flags | Text | P2P=auto, VLLM_ENGRAM_PLACEMENT=auto (ram or ssd), MAX_SEQS 8, MAX_BATCHED 2048, DSpark drafter up to 7 tokens | |
| status | String | Proven · Reported · Theory · Contested · Warning | Reported |
| tested_by | String | voidgiraffe | |
| tested_on | Date | 2026-10-24 | |
| source | String | https://github.com/Morrowmake/dsv41-flash-cmp170hx-recipe | |
| files | Wikitext |