Bizuayeu/GLM-5.3-Flash-XSFP4

GLM-5.3-Flash XSFP4

A derived checkpoint of nvidia/GLM-5.3-Flash-FP4-Local (revision 76348ddb2be2e35e39ecfc00e75dc07703f4f3d5). The routed experts of layers 3 to 44 and the MTP draft layer's experts are requantized from the BF16 weights of zai-org/GLM-5.3-Flash as XSFP4, a standard NVFP4 layout with no runtime smoothing; every other tensor is FP4-Local's, byte for byte. It is not an NVIDIA or Z.AI release. It is served by the 3.x line of NVFP4 BIZ (TensorFold's native engine, from 3.3.0) on GB10 (DGX Spark class) systems.

Status. Published ahead of NVFP4 BIZ 3.3.0, the release that serves it, which is in its acceptance; that release's serving measurements will be added here when it is out.

At a glance

Source	nvidia/GLM-5.3-Flash-FP4-Local, revision 76348ddb2be2e35e39ecfc00e75dc07703f4f3d5 (NVIDIA Open Model Agreement)
Upstream	zai-org/GLM-5.3-Flash (MIT, Copyright (c) 2026 Z.AI Co., Ltd)
Requantized	Routed experts of layers 3 to 44 and the MTP layer's experts: XSFP4 (below), NVFP4 weights with FP8 E4M3 block scales of 16, from the BF16 weights
As shipped by FP4-Local	KDA projections, dense MLP and lm_head (MXFP8); MLA and shared experts (FP8, per block); router; norms (F32); embeddings; the vision tower
Calibration	512 rows of 2,048 tokens: Japanese 35% (Nemotron-Post-Training-Dataset-v2 Japanese, LLM-jp), local work logs 25%, math 15%, STEM 10%, English conversation 15%; rendered with the chat template at reasoning effort high; routing-only (each expert sees only the tokens routed to it)
Files	186 GB: 32 shards of the routed experts (model-xsfp4experts-*), one of the MTP layer's experts (model-mtp-xsfp4experts.safetensors), FP4-Local's other tensors in model-0000{1..5}-of-00040.safetensors and tensors/native-QB.safetensors, the index, and the source's config.json, hf_quant_config.json, generation_config.json, processor_config.json, tokenizer and chat template; NVIDIA's model card kept as README.nvidia.md
License	NVIDIA Open Model Agreement (this derivative; LICENSE, NOTICE); the upstream model is MIT

XSFP4

XSFP4 folds OSFP4's smoothing into the weights: for each expert, the smoothing of the down projection's input is multiplied into the rows of the up projection, and the gate and up projections take no smoothing of their own ("plan 2"). The block scales are then fitted again, so the stored weights are ordinary NVFP4 that any NVFP4 kernel reads, with nothing to apply at run time.

Measured

Each checkpoint's next-token distributions over the same 128 documents (144,285 positions) against those of the BF16 weights.

On the engine that serves it (TensorFold's native engine as NVFP4 BIZ 3.3.0 builds it: upstream 1.1.1 with upstream pull request #1050, which loads FP4-Local's formats as stored; two GB10 systems at TP=2), each document sent as token ids with prompt_logprobs of 20:

Checkpoint	Top-1 agreement	NLL above BF16	KL, lower bound
This checkpoint	93.26%	+0.85%	0.0431
nvidia/GLM-5.3-Flash-FP4-Local	92.82%	+1.11%	0.0510
nvidia/GLM-5.3-Flash-NVFP4	92.69%	+1.37%	0.0510

This checkpoint is closer to BF16 than each of the other two by all three measures, the per-document 95% intervals of the differences holding no zero (against FP4-Local's NLL narrowly: +0.01 to +0.55 points). FP4-Local and NVIDIA's NVFP4 are not told apart by the KL bound or the NLL. The KL bound counts one by one the tokens whose probability both sides report (the engine's best 20, BF16's best 20 and the text's token) and the rest as one bucket: it is exact from these numbers and lower than the KL over the whole vocabulary.

Outside the engine, the same documents through the model in PyTorch with each checkpoint's weights dequantized and 16-bit activations (W4A16), over the whole vocabulary (these KL values do not compare with the bound above):

Checkpoint	KL	Top-1 agreement	NLL above BF16
This checkpoint (the MTP layer does not enter the next token)	0.0410	93.90%	+1.23%
nvidia/GLM-5.3-Flash-FP4-Local	0.0528	93.17%	+1.58%
nvidia/GLM-5.3-Flash-NVFP4	0.0512	93.22%	+1.99%

The MTP layer, which drafts tokens, was compared on the same 128 documents (144,157 positions) with FP4-Local's own MTP layer, teacher-forced at depth 1: its draft agreed with the main model's greedy token at 80.273% against 80.295%, a difference of −0.022 points with a per-document 95% interval of [−0.082, +0.036], and with the text's token two ahead at 68.607% against 68.636% ([−0.078, +0.022]). Both intervals hold zero, so the two cannot be told apart there, and the requantized layer ships.

License

This checkpoint is a Derivative Work of nvidia/GLM-5.3-Flash-FP4-Local and is distributed under the NVIDIA Open Model Agreement, a copy of which is in LICENSE. Licensed by NVIDIA Corporation under the NVIDIA Open Model Agreement. The GLM-5.3-Flash base model is governed by the MIT License (Copyright (c) 2026 Z.AI Co., Ltd). The calibration data's Nemotron-Post-Training-Dataset-v2 rows are CC-BY-4.0 (NVIDIA); the data itself is not distributed.