# Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 by Ethan Todd Source: https://savrn.com/models/swift-1-5-qwen3-8-27b-quark-rtn-mxfp4 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 (15.6B 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 | 31.2 GB | 37.5 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 | | 8-bit | 15.6 GB | 18.7 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 | | 4-bit | 7.8 GB | 9.4 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 7, 2026. [Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/swift-1-5-qwen3-8-27b-quark-rtn-mxfp4/gpus) ## Model Card quantized to OCP MXFP4 4-bit weights in the same Quark checkpoint container as The weights are plain RTN, not AWQ: see the note below. The 15 MTP tensors are kept in BF16; the radiance runtime loads them with RADIANCEQUARKBF16MTP=1. Why RTN instead of AWQ. The first build of this checkpoint used AWQ with the same smoothing recipe as AMD's release. The AWQ fold itself was mathematically consistent, but on this model a few layers converged to weights to 32-element MXFP4 blocks then destroyed those layers on real, outlier-carrying inputs (layer 7 output relative error ~324), and repairing the worst layers individually was not enough; the remaining smoothed layers still accumulated too much… Excerpt from the card by Ethan Todd, licensed other. ## Configuration Architecture Qwen3_5ForConditionalGeneration Context length (tokens) 262,144 Layers 64 Hidden size 5,120 Feed-forward size 17,408 Attention heads 24 Key/value heads 4 Head dimension 256 Vocabulary size 248,320 Model type qwen3_5 Quantization quark ## Identity and Version Repository ethantodd4l/Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 Publisher Ethan Todd Task Image and text to text Modality Image and text Library transformers Parameters 15.6B parameters Languages Not stated by the source Revision 3cd117e0d4b1499324bf92479cf2ca30b463041e First published 2026-10-03 Last updated 2026-10-04 ## Files and Weights 14 files, 19.8 GB in total. The weights are 1 file totalling 19.8 GB in safetensors. Weights1 file · 19.8 GB Configuration4 files · 12.0 KB Tokenizer4 files · 22.9 MB Documentation3 files · 21.9 KB Other1 file · 9.0 KB Repository1 file · 1.7 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model.safetensors | Weights | 19.8 GB | 74b8f551511d | | config.json | Configuration | 11.0 KB | — | | generation_config.json | Configuration | 214 B | — | | preprocessor_config.json | Configuration | 390 B | — | | video_preprocessor_config.json | Configuration | 385 B | — | | LICENSE | Documentation | 13.3 KB | — | | NOTICE | Documentation | 1.5 KB | — | | README.md | Documentation | 7.0 KB | — | | chat_template.jinja | Other | 9.0 KB | — | | .gitattributes | Repository | 1.7 KB | — | | merges.txt | Tokenizer | 3.4 MB | — | | tokenizer.json | Tokenizer | 12.8 MB | 0997f410c57a | | tokenizer_config.json | Tokenizer | 17.9 KB | — | | vocab.json | Tokenizer | 6.7 MB | — | ## License and Download License other Access Open weights, no gate Download size 19.8 GB [Download from Ethan Todd](https://huggingface.co/ethantodd4l/Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4) Released by Ethan Todd through its official repository on Hugging Face. [Read the license](https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27b/blob/main/LICENSE). ## Built From - Derived from ukisai/Swift-1.5-Qwen3.8-27b - Quantized from ukisai/Swift-1.5-Qwen3.8-27b ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 19.8 GB | | 16-bit | 31.2 GB | | 8-bit | 15.6 GB | | 4-bit | 7.8 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Questions About Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 ### How much GPU memory does Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 need? About 37.5 GB at 16-bit and 9.4 GB at 4-bit: the weights (15.6B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 on? At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists. ### What license is Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4 released under? other, as its publisher declares it. Read the license text before commercial use. ### What is Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4's context length? 262,144 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Image and text to text ### [Qwen3.8-27B-NVFP4-RTX5090](https://savrn.com/models/qwen3-8-27b-nvfp4-rtx5090) [Gittensor Model Hub](https://savrn.com/model-publishers/gittensor-model-hub) Runs on SparkInfer, SGLang and vLLM unmodified — configs for all three are below. Serving many users at once? See concurrency. SparkInfer × this NVFP4 build × the DSpark v2 drafter — an engine, a checkpoint, and a speculative drafter optimized against each other, compounding to 4.3×. The drafter never changes what the model says: the target verifies every drafted token. GeForce RTX 5090–specific NVFP4 checkpoint of Qwen/Qwen3.8-27B, quantized with NVIDIA Model Optimizer. Serves the full native 262,144-token context on 32 GB. With the DSpark v2 drafter: 264.8 tok/s overall — up to 420 on code — on SparkInfer (its bench harness; the HTTP server is autoregressive-only today) and 161.7 tok/s on… Open weights apache-2.0 14.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-8-27b-nvfp4-rtx5090) Model · Image and text to text ### [GLM-5.3-Flash-NestQuant-1.5-4bit](https://savrn.com/models/glm-5-3-flash-nestquant-1-5-4bit) [Jarrel Seah](https://savrn.com/model-publishers/jarrelscy) GLM-5.3-Flash with its routed experts quantized to NestQuant, a nested two-level format. Each expert has a 1.5-bit base plus a residual plane that lifts it to 4 bits. At serving time an expert moves from 1.5 bit to 4 bit by loading its residual plane on top of the base, and the base bytes stay the same. This build targets a 128 GB Mac with 96 GB usable for weights, KV cache and runtime. The native vision tower is included, so the model still takes images. Status: TODO. Encoding in progress. This repo cannot be loaded with stock transformers, vLLM or MLX. The serving kernel and loader will be published separately. - 67 of 288 experts per layer (23%) run at 4 bit. The rest run at 1.5 bit.… Open weights mit 16.9B parameters 1,048,576 tokens [View model](https://savrn.com/models/glm-5-3-flash-nestquant-1-5-4bit) Model · Image and text to text ### [Qwen3.8-27B-NVFP4](https://savrn.com/models/qwen3-8-27b-nvfp4-3) [Inferact Inc.](https://savrn.com/model-publishers/inferact) Quantized version of https://huggingface.co/Qwen/Qwen3.8-27B Open weights apache-2.0 17.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-8-27b-nvfp4-3) Model · Image and text to text ### [Qwen3.8-27B-NVFP4](https://savrn.com/models/qwen3-8-27b-nvfp4-2) [RadixArk](https://savrn.com/model-publishers/radixark) The RadixArk Qwen3.8-27B-NVFP4 model is the quantized version of Qwen/Qwen3.8-27B. The quantization was produced at RadixArk using NVIDIA Model Optimizer, following a mixed NVFP4 W4A4 recipe. Run on SGLang: launch command and per-platform recipes in the Qwen3.8-27B cookbook. This model is not owned or developed by RadixArk. It is a quantized derivative of Qwen's model; see the upstream Qwen3.8-27B model card for the source model's capabilities, training information, limitations, and license. Global Developers looking to deploy an off-the-shelf, pre-quantized model in AI agent systems, chatbots, RAG systems, and other AI-powered applications. Hugging Face 08/14/2026 via… Open weights apache-2.0 18.2B parameters 262,144 tokens Model Optimizer [View model](https://savrn.com/models/qwen3-8-27b-nvfp4-2) Model · Image and text to text ### [Swift-Qwen3.8-27B-Uncensored-NVFP4](https://savrn.com/models/swift-qwen3-8-27b-uncensored-nvfp4) [AJ Gazin](https://savrn.com/model-publishers/ajgazin) NVFP4 checkpoint of an abliterated Swift-Qwen3.8-27B (UkisAI's reasoning-efficient fine-tune of Qwen3.8-27B). For vLLM and SGLang. GGUFs for llama.cpp: The Swift 1.5 version is source). - Swift's own NVFP4 recipe, unmodified, from ukisai/Swift-Qwen3.8-27B-NVFP4, calibrated with NVIDIA ModelOpt. - MTP head and vision tower in BF16, bit-identical to the source. 21.9 GB, NVIDIA ModelOpt mixed-precision format. Needs a vLLM with ModelOpt mixed-precision support (tested on 0.29.0). No --quantization flag. Sampling, as for Swift and Qwen: temperature 1.0, topp 0.95, topk 20, minp 0. The model thinks before answering by default. Tested on an RTX 5090 (32 GB) with vLLM 0.29.0: NVFP4 layers on… Open weights other 18.2B parameters 262,144 tokens vllm [View model](https://savrn.com/models/swift-qwen3-8-27b-uncensored-nvfp4) Model · Image and text to text ### [Swift-1.5-Qwen3.8-27B-Uncensored-NVFP4](https://savrn.com/models/swift-1-5-qwen3-8-27b-uncensored-nvfp4) [AJ Gazin](https://savrn.com/model-publishers/ajgazin) NVFP4 checkpoint of an abliterated Swift 1.5 Qwen3.8-27B (UkisAI's reasoning-efficient fine-tune of Qwen3.8-27B). For vLLM and SGLang. GGUFs for llama.cpp: (measured on the BF16 source). - Swift's own NVFP4 recipe, unmodified, from ukisai/Swift-Qwen3.8-27B-NVFP4, calibrated with NVIDIA ModelOpt. The module split matches UkisAI's Swift 1.5 NVFP4 exactly. - MTP head and vision tower in BF16, bit-identical to the source. 21.9 GB, NVIDIA ModelOpt mixed-precision format. Needs a vLLM with ModelOpt mixed-precision support. No --quantization flag. Sampling, as for Swift and Qwen: temperature 1.0, topp 0.95, topk 20, minp 0. The model thinks before answering by default. Same format, recipe, module… Open weights other 18.2B parameters 262,144 tokens vllm [View model](https://savrn.com/models/swift-1-5-qwen3-8-27b-uncensored-nvfp4) ## Ethan Todd [All models and datasets](https://savrn.com/model-publishers/ethantodd4l) ## Versions - [3cd117e0d4b1](https://savrn.com/models/swift-1-5-qwen3-8-27b-quark-rtn-mxfp4/versions/3cd117e0d4b1) · current 2026-10-04 ## Explore More - [All image and text to text models](https://savrn.com/models/tasks/image-and-text-to-text) - [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 2026-10-04. - [Hugging Face record](https://huggingface.co/ethantodd4l/Swift-1.5-Qwen3.8-27b-Quark-RTN-MXFP4) - [How the hub is built](https://savrn.com/model-hub/methodology)