# GLM5.3-Flash-E224-DGX-Spark by AutoTrust AI Lab: Open Model Source: https://savrn.com/models/glm5-3-flash-e224-dgx-spark Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve GLM5.3-Flash-E224-DGX-Spark (127.6B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them. | Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits | | --- | --- | --- | --- | --- | --- | | 16-bit | 255.3 GB | 306.3 GB | 2x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $3.70 | [2x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $4.00 · [2x MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) $5.18 | | 8-bit | 127.6 GB | 153.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 · [1x MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) $2.59 | | 4-bit | 63.8 GB | 76.6 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 | Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 12, 2026. [GLM5.3-Flash-E224-DGX-Spark on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/glm5-3-flash-e224-dgx-spark/gpus) ## Model Card By AutoTrust AI Lab, published under mit, revision 593eb0dfa230. autotrust/GLM5.3-Flash-E224-DGX-Spark is a compact build of [zai-org/GLM-5.3-Flash](https://savrn.com/models/glm-5-3-flash) for desktop Blackwell systems like NVIDIA DGX Spark. It's an unofficial derivative. Deploying on DGX Spark? A complete, field-tested runbook for serving this checkpoint on two ConnectX-7-linked DGX Sparks (vLLM 0.31.0, TP=2, fp8 KV cache) is here: DEPLOY-2X-DGX-SPARK.md — network setup, NCCL/CX7 pitfalls, memory budgeting, ops & troubleshooting, and measured throughput (15.7 tok/s single-stream, 71.8 tok/s at 8-way concurrency, both GPUs ≈96% utilized). It keeps 224 of the 288 routed experts in each layer by Neural Architecture Search (NAS), uses NVFP4 for the experts and still activates 18 B parameters per token. The weights take 141 GiB, which is small enough for two DGX Sparks connected by ConnectX-7 (256 GB of unified memory in total) with room left for long-context KV cache. A single 180 GB Blackwell GPU (B200/GB200) can also run it. [Read the full model card (2,295 words)](https://savrn.com/models/glm5-3-flash-e224-dgx-spark/card) ## Configuration Architecture Glm5NextForConditionalGeneration Context length (tokens) 1,048,576 Layers 45 Hidden size 4,096 Feed-forward size 12,288 Attention heads 64 Key/value heads 64 Head dimension 0 Vocabulary size 154,880 Routed experts 224 Experts active per token 8 Model type glm5_next Quantization modelopt ## Identity and Version Repository autotrust/GLM5.3-Flash-E224-DGX-Spark Publisher AutoTrust AI Lab Task Image and text to text Modality Image and text Library vllm Parameters 127.6B parameters Languages en, zh Revision 593eb0dfa23092ac932cd87897f80d3c8e43beb2 First published 2026-10-06 Last updated 2026-10-10 ## Files and Weights 64 files, 175.9 GB in total. The weights are 34 files totalling 175.8 GB in safetensors. Weights34 files · 175.8 GB Configuration17 files · 12.8 MB Tokenizer6 files · 60.7 MB Documentation3 files · 66.3 KB Other3 files · 31.9 KB Repository1 file · 1.6 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model-00001-of-00029.safetensors | Weights | 5.4 GB | 934308d2ec44 | | model-00002-of-00029.safetensors | Weights | 4.3 GB | 35020d79d4bf | | model-00003-of-00029.safetensors | Weights | 5.4 GB | 7eff91d0a951 | | model-00004-of-00029.safetensors | Weights | 5.4 GB | 0994af1d3926 | | model-00005-of-00029.safetensors | Weights | 5.4 GB | 8a541de81b94 | | model-00006-of-00029.safetensors | Weights | 5.4 GB | 512417457494 | | model-00007-of-00029.safetensors | Weights | 5.4 GB | 80e4ba208d6e | | model-00008-of-00029.safetensors | Weights | 5.4 GB | 43de26b8b418 | | model-00009-of-00029.safetensors | Weights | 5.4 GB | 438d51410b7d | | model-00010-of-00029.safetensors | Weights | 5.4 GB | b7de6a1a3237 | | model-00011-of-00029.safetensors | Weights | 5.4 GB | 2d0cc5d36928 | | model-00012-of-00029.safetensors | Weights | 5.3 GB | 844feb3b6883 | | model-00013-of-00029.safetensors | Weights | 5.4 GB | 48b81c43881d | | model-00014-of-00029.safetensors | Weights | 5.4 GB | b3504234be9c | | model-00015-of-00029.safetensors | Weights | 5.4 GB | 96187f0b9798 | | model-00016-of-00029.safetensors | Weights | 5.4 GB | b23785d9342f | | model-00017-of-00029.safetensors | Weights | 5.4 GB | 1d4af87ea7cc | | model-00018-of-00029.safetensors | Weights | 5.4 GB | 36e91f213ee4 | | model-00019-of-00029.safetensors | Weights | 5.4 GB | 1c8f3baad732 | | model-00020-of-00029.safetensors | Weights | 5.4 GB | fa2d13b3ceed | | model-00021-of-00029.safetensors | Weights | 5.4 GB | 930aba37b157 | | model-00022-of-00029.safetensors | Weights | 5.3 GB | 86e74e04c21c | | model-00023-of-00029.safetensors | Weights | 5.4 GB | c7fbc30c080e | | model-00024-of-00029.safetensors | Weights | 5.4 GB | fe24a9a07aa3 | | model-00025-of-00029.safetensors | Weights | 5.4 GB | 60bb191fad97 | | model-00026-of-00029.safetensors | Weights | 5.4 GB | 859b97856dcf | | model-00027-of-00029.safetensors | Weights | 5.4 GB | 344632994228 | | model-00028-of-00029.safetensors | Weights | 5.4 GB | a2ef0a5d5109 | | model-00029-of-00029.safetensors | Weights | 2.4 GB | e9331b1e676e | | mtp-nvfp4/model.safetensors | Weights | 6.9 GB | 7d75ea350cd7 | | mtp/model-00001-of-00004.safetensors | Weights | 5.4 GB | 6ea3c5423028 | | mtp/model-00002-of-00004.safetensors | Weights | 5.4 GB | a68fe8edc360 | | mtp/model-00003-of-00004.safetensors | Weights | 5.4 GB | df4412b08e69 | | mtp/model-00004-of-00004.safetensors | Weights | 1.3 GB | adaae846c8e0 | | config.json | Configuration | 13.9 KB | — | | generation_config.json | Configuration | 215 B | — | | hf_quant_config.json | Configuration | 7.2 KB | — | | model.safetensors.index.json | Configuration | 12.6 MB | 0cd233a76dd9 | | mtp-nvfp4/config.json | Configuration | 7.3 KB | — | | mtp-nvfp4/generation_config.json | Configuration | 215 B | — | | mtp-nvfp4/hf_quant_config.json | Configuration | 1.0 KB | — | | mtp-nvfp4/processor_config.json | Configuration | 909 B | — | | mtp/config.json | Configuration | 13.9 KB | — | | mtp/generation_config.json | Configuration | 215 B | — | | mtp/hf_quant_config.json | Configuration | 7.2 KB | — | | mtp/model.safetensors.index.json | Configuration | 94.4 KB | 4e2e393f9fea | | mtp/processor_config.json | Configuration | 909 B | — | | processor_config.json | Configuration | 909 B | — | | scripts/export_mtp_nvfp4.py | Configuration | 6.0 KB | — | | scripts/quantize_mtp_nvfp4.py | Configuration | 6.8 KB | — | | scripts/validate_nvfp4_format.py | Configuration | 4.1 KB | — | | DEPLOY-2X-DGX-SPARK.md | Documentation | 37.3 KB | — | | MTP-NVFP4.md | Documentation | 12.0 KB | — | | README.md | Documentation | 17.0 KB | — | | chat_template.jinja | Other | 10.6 KB | — | | mtp-nvfp4/chat_template.jinja | Other | 10.6 KB | — | | mtp/chat_template.jinja | Other | 10.6 KB | — | | .gitattributes | Repository | 1.6 KB | — | | mtp-nvfp4/tokenizer.json | Tokenizer | 20.2 MB | 19e773648cb4 | | mtp-nvfp4/tokenizer_config.json | Tokenizer | 761 B | — | | mtp/tokenizer.json | Tokenizer | 20.2 MB | 19e773648cb4 | | mtp/tokenizer_config.json | Tokenizer | 761 B | — | | tokenizer.json | Tokenizer | 20.2 MB | 19e773648cb4 | | tokenizer_config.json | Tokenizer | 761 B | — | ## License and Download License mit Access Open weights, no gate Download size 175.8 GB [Download from AutoTrust AI Lab](https://huggingface.co/autotrust/GLM5.3-Flash-E224-DGX-Spark) Released by AutoTrust AI Lab through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Derived from [zai-org/GLM-5.3-Flash](https://savrn.com/models/glm-5-3-flash) - Quantized from [zai-org/GLM-5.3-Flash](https://savrn.com/models/glm-5-3-flash) ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 175.8 GB | | 16-bit | 255.3 GB | | 8-bit | 127.6 GB | | 4-bit | 63.8 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About GLM5.3-Flash-E224-DGX-Spark ### How much GPU memory does GLM5.3-Flash-E224-DGX-Spark need? About 306.3 GB at 16-bit and 76.6 GB at 4-bit: the weights (127.6B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run GLM5.3-Flash-E224-DGX-Spark on? At 16-bit, 2x MI300X from $3.70 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists. ### Can I use GLM5.3-Flash-E224-DGX-Spark commercially? Yes. GLM5.3-Flash-E224-DGX-Spark is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. ### What is GLM5.3-Flash-E224-DGX-Spark's context length? 1,048,576 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Image and text to text ### [Qwen3.5-122B-A10B-FP8](https://savrn.com/models/qwen3-5-122b-a10b-fp8) [Qwen](https://savrn.com/model-publishers/qwen) Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance. Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility to empower developers and enterprises with unprecedented capability and efficiency. For more details, please refer to our blog post Qwen3.5. WMT24++: a harder subset of WMT24 after difficulty labeling and rebalancing; we report the averaged scores on 55 languages using XCOMET-XXL. Empty cells (--) indicate scores not yet available or not applicable. Empty cells (--) indicate scores not… Open weights apache-2.0 125.1B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-5-122b-a10b-fp8) Model · Image and text to text ### [tinfield-1-nvfp4-probe](https://savrn.com/models/tinfield-1-nvfp4-probe) [Natapong Nitarach (Schwyter)](https://savrn.com/model-publishers/natnitaract) Public research conversion of badtheorylabs/Tinfield-1, derived from Qwen/Qwen3.8-Flash-Next. The source checkpoint was pinned to f7c714a97b71342db9e2fc04009d2704c656c80c. This Hugging Face repository contains the complete original export, including all 10 safetensors shards: - PLE embeddings, vision, MTP and other non-expert weights: retained in BF16. - The original configuration, tokenizer, processors and tensor index accompany the shards. The ple-int8-e579 Kaggle variant is a separate quantized PLE derivative. This Hugging Face checkpoint contains BF16 PLE. Load all shards and the configuration from this repository together. Conversion used NVIDIA Model-Optimizer recipe commit… Open weights other 117B parameters 262,144 tokens transformers [View model](https://savrn.com/models/tinfield-1-nvfp4-probe) Model · Image and text to text ### [GLM5.3-Flash-E256-DGX-Spark](https://savrn.com/models/glm5-3-flash-e256-dgx-spark) [AutoTrust AI Lab](https://savrn.com/model-publishers/autotrust) autotrust/GLM5.3-Flash-E256-DGX-Spark is a compact build of zai-org/GLM-5.3-Flash for Blackwell systems, from DGX Spark clusters to a single B200. It's an unofficial derivative. It keeps 256 of the 288 routed experts in each layer by Neural Architecture Search (NAS), uses NVFP4 for the experts and still activates 18 B parameters per token. It's the largest architecture in the family, with the best Chinese, code and vision scores. The KV cache uses FP8. The repo includes a ready-to-use NVFP4 MTP draft in mtp-nvfp4/ (6.95 GB) for speculative decoding. On a single B200 it gives 1.84× single-stream decode speed (1.94× with 3 draft tokens), and output quality is unchanged. The model keeps the… Open weights mit 144.5B parameters 1,048,576 tokens vllm [View model](https://savrn.com/models/glm5-3-flash-e256-dgx-spark) Model · Image and text to text ### [GLM5.3-Flash-E192-DGX-Spark](https://savrn.com/models/glm5-3-flash-e192-dgx-spark) [AutoTrust AI Lab](https://savrn.com/model-publishers/autotrust) autotrust/GLM5.3-Flash-E192-DGX-Spark is a compact build of zai-org/GLM-5.3-Flash for desktop Blackwell systems like NVIDIA DGX Spark. It's an unofficial derivative. It keeps 192 of the 288 routed experts in each layer by Neural Architecture Search (NAS), uses NVFP4 for the experts and still activates 18 B parameters per token. The weights take 124 GiB, the smallest in the family: two DGX Sparks connected by ConnectX-7 get about 62 GiB of weights per node and the most KV room of any build, and a single 180 GB Blackwell GPU (B200/GB200) runs it with 32 GiB of KV cache. The KV cache uses FP8. The repo includes a ready-to-use NVFP4 MTP draft in mtp-nvfp4/ (6.95 GB) for speculative decoding. On… Open weights mit 110.7B parameters 1,048,576 tokens vllm [View model](https://savrn.com/models/glm5-3-flash-e192-dgx-spark) Model · Image and text to text ### [clover-1-150b-preview](https://savrn.com/models/clover-1-150b-preview) [Kitani](https://savrn.com/model-publishers/kitaniai) libraryname: transformers pipelinetag: image-text-to-text basemodel: - Qwen/Qwen3.8-27B - clover - mixture-of-experts - multimodal - reasoning - open-weights - custom-code Clover 1 150B Preview is the first public checkpoint of Clover 1, an experimental MoE model we're developing at Kitani. It's based on Qwen3.8-27B, but it isn't just a finetune with a different name. We kept much of Qwen3.8's underlying architecture, including the tokenizer, multimodal components, hybrid language backbone, embeddings, and LM head, while replacing the language model's dense FFNs with our own sparse mixture-of-experts setup. Clover currently has 8 full-sized experts per language layer with top-1 routing. The… Open weights apache-2.0 147.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/clover-1-150b-preview) Model · Image and text to text ### [DeepSeek-V4.1-Flash-oQ2.5e-mtp](https://savrn.com/models/deepseek-v4-1-flash-oq2-5e-mtp) [Jun Kim](https://savrn.com/model-publishers/jundot) This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385). A mixed-precision MLX checkpoint of deepseek-ai/DeepSeek-V4.1-Flash, a multimodal mixture-of-experts model with 552B backbone parameters, a Causal Encoder-Decoder layout, Compressed Sparse Attention 2 (CSA2), Engram conditional memory and DSpark speculative decoding. It includes the language model, the Engram tables, the vision encoder, the multimodal aligner and the DSpark MTP weights. The effective average is 3.104 bits/weight excluding the Engram tables, or 3.592 bits/weight for the complete checkpoint. oQ2.5e allocates bits by measured layer sensitivity, then… Open weights mit 102.4B parameters 1,048,576 tokens mlx [View model](https://savrn.com/models/deepseek-v4-1-flash-oq2-5e-mtp) ## AutoTrust AI Lab [All models and datasets](https://savrn.com/model-publishers/autotrust) ## Versions - [593eb0dfa230](https://savrn.com/models/glm5-3-flash-e224-dgx-spark/versions/593eb0dfa230) · current 2026-10-10 ## Explore More - [All image and text to text models](https://savrn.com/models/tasks/image-and-text-to-text) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-10. - [Hugging Face record](https://huggingface.co/autotrust/GLM5.3-Flash-E224-DGX-Spark) - [How the hub is built](https://savrn.com/model-hub/methodology)