GLM-5.3-Flash-DFlash2-MXFP8
This repository contains an MXFP8-quantized DFlash 2 draft model for
local-inference-lab/GLM-5.3-Flash-NVFP4.
It is not a standalone language model. A compatible speculative-decoding
server loads it beside the target model and verifies every drafted token
against the target.
The source checkpoint is
incoai/GLM-5.3-Flash-DFlash2
at immutable revision dc77ff1c99eeb2df044ee3d4f0094eb033fee410.
Format
- Linear weights:
float8_e4m3fn - Scale values: biased E8M0 exponents stored as
uint8 - Quantization block: 1×32 values
- Scale layout: row-major and unswizzled
- Excluded module:
lm_head - Draft KV cache quantization: not encoded in the checkpoint
conversion_manifest.json records the immutable source revision, source and
output checksums, tensor coverage, aggregate quantization error, and
per-weight validation statistics.
Validation status
Status: qualified for checkpoint structure, exact format reproduction, loading, and smoke inference under the following conditions:
- Target:
local-inference-lab/GLM-5.3-Flash-NVFP4revision520de24eabf507659eaef7c70f14fd584527facc - Runtime:
voipmonitor/vllm:glm53-flash-nvfp4-dflash2-community-20260830-r5 - Hardware: four NVIDIA RTX PRO 6000 Blackwell Workstation Edition GPUs
- Parallelism: tensor parallel size 4 and decode-context parallel size 1
- Target attention, MoE, linear, and tensor-parallel all-reduce: B12X
- DFlash attention: FlashAttention 2
- DFlash linear: B12X MXFP8
- DFlash proposal length: seven tokens
- DFlash KV cache:
auto(BF16) - CUDA graph mode:
NONEfor the load-and-inference qualification
The runtime detected ModelOpt MXFP8, selected B12xMxfp8LinearKernel for
draft GEMMs and the fused DFlash context K/V projection, loaded 1.20 GB of
draft weights, and completed a 32-token speculative request. Runtime
throughput and CUDA graph performance are not qualified by that smoke test.
The checkpoint is unsupported in vLLM builds that do not contain the DFlash 2 and ModelOpt MXFP8 integration used by the qualified runtime.
Serving
docker run --rm \
--gpus '"device=0,1,2,3"' \
--network host \
--ipc host \
--shm-size 32g \
-e MODEL=local-inference-lab/GLM-5.3-Flash-NVFP4 \
-e SERVED_MODEL_NAME=GLM-5.3-Flash-NVFP4 \
-e PORT=8000 \
-e TP=4 \
-e DCP=1 \
-e MAX_NUM_SEQS=16 \
-e MAX_MODEL_LEN=262144 \
-e MAX_NUM_BATCHED_TOKENS=4096 \
-e SPECULATOR=dflash \
-e NUM_SPECULATIVE_TOKENS=7 \
-e DFLASH_MODEL=local-inference-lab/GLM-5.3-Flash-DFlash2-MXFP8 \
-e DFLASH_MODEL_REVISION= \
-e DFLASH_KV_CACHE_DTYPE=auto \
-e DFLASH_ATTENTION_BACKEND=FLASH_ATTN \
-e ATTENTION_BACKEND=B12X \
-e MOE_BACKEND=b12x \
-e LINEAR_BACKEND=b12x \
-e B12X_PCIE_ALLREDUCE=1 \
-e CUDAGRAPH_MODE=FULL \
-e VLLM_B12X_MOE_FP4_FORCE_A16=0 \
voipmonitor/vllm:glm53-flash-nvfp4-dflash2-community-20260830-r5
An empty DFLASH_MODEL_REVISION makes the launcher resolve the repository's
main branch. For reproducible deployments, replace the empty value with an
immutable Hugging Face commit hash. The OpenAI-compatible endpoint is
available at http://127.0.0.1:8000/v1.
License and attribution
The source DFlash 2 model is distributed under CC BY-NC-ND 4.0. See the source model card for its use restrictions and attribution information.
@misc{inco2026dflash2,
title = {{DFlash 2: Keep Drafting Parallel}},
author = {{Inco AI}},
year = {2026},
month = {August},
url = {https://inco.ai/blog/dflash2/}
}