Page values for "GLM 5.3 on 8x 8 GB (Int4-Int8Mix PP8)"
From CMP Wiki
More actions
"Recipe" values
1 row is stored for this page| Field | Field type | Allowed values | Value |
|---|---|---|---|
| page | Page | GLM 5.3 on 8x 8 GB (Int4-Int8Mix PP8) | |
| model | String | GLM-5.3 (full 78-layer, not Flash) | |
| quant | String | GLM-5.3-Int4-Int8Mix (INT4 / INT8 mix); KV fp8_ds_mla | |
| engine | String | vLLM (build not stated) | |
| engine_version | String | ||
| cards | String | 8 GB · 10 GB · both | 8 GB |
| card_count | Integer | 8 | |
| topology | String | PP8, TP1 | |
| context | String | 1,048,576 | |
| decode_tps | Float | 25.0 | |
| prefill_tps | Float | ||
| vram_per_card | String | 64 GiB, --gpu-memory-utilization 0.97 | |
| power_cap | String | 200 W | |
| driver | String | ||
| unlock_build | String | ||
| flags | Text | --pipeline-parallel-size 8 --tensor-parallel-size 1 --speculative-config mtp 5 --attention-backend TRITON_MLA_SPARSE --kv-cache-dtype fp8_ds_mla --block-size 64 --no-async-scheduling | |
| status | String | Proven · Reported · Theory · Contested · Warning | Reported |
| tested_by | String | 1a2b3c0967 | |
| tested_on | Date | 2026-10-07 | |
| source | String | https://discord.com/channels/1518634379604136178/1528826330454429706/1557523316120748073 | |
| files | Wikitext |