ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF · Hugging Face

Source: https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#swift-15-qwen38-27b-%C2%B7-gsq-rcoSwift 1.5 Qwen3.8-27B · GSQ-RCO

Compact, mixed-precision GGUF quantizations ofSwift 1.5 Qwen3.8-27B, with Swift-specific refinement using the per-tensor allocations fromISTA-DASLab’s GSQ-RCO release.

Swift 1.5 builds on Swift 1.0 through post-training focused on long-horizon, agentic and coding tasks, improving overall performance while using fewer thinking tokens. See theoriginal model cardfor the model’s training approach and benchmark results. Those model-level benchmarks are separate from the quantization measurements below.

Swift 1.5 uses58.5% fewer thinking tokenswhile scoring0.35% higherthan the base, for a9.18× speed-upon several tasks.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#available-quantizationsAvailable quantizations

Each tier is a single GGUF file. Sizes are decimal GB; runtime memory also includes the context cache and compute buffers. Tier names denote mixed-precision allocation profiles, rather than a uniform type for every tensor.

The optional\-mtpfiles retain the matching refined model tensors and add the MTP head. They require a runtime with support for this model’s MTP implementation. The KLD results here were measured on thestandard files; MTP decoding speed and quality have not been separately evaluated.

Exact file identities are recorded inrelease-manifest.jsonandSHA256SUMS.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#evaluationEvaluation

KLD measures the divergence of the quantized model’s next-token distribution fromSwift 1.5 BF16; lower is better. Development measurements usewiki\.test\.raw, 100 chunks and a512-token context. Development data informed refinement and is not independent validation.

The held-out results below use C4 prose, CodeParrot code, GSM8K math text and multilingual mC4 text. Prose/code/math use 100 chunks each; German/French/Spanish/Chinese use 25 chunks each, all at context 512. These are distributional KLD measurements, not task accuracy or math benchmark scores.

Held-out textIQ2_XSIQ2_SIQ3_XXSIQ3_SC4 prose0.1615940.1074380.0802380.041748CodeParrot code0.1208840.0847390.0625460.035447GSM8K math text0.1174670.0963480.0759640.043916German0.1244820.0914220.0746450.035698French0.1667850.1132770.0777910.046219Spanish0.0823130.0563930.0399060.024186Chinese0.2070630.1294850.1031410.052241 All four refined files improve KLD over their matched Swift starting quantizations on all seven reporting domains. Results are not uniformly better than the ISTA comparison quants: math-text KLD is 4.9–7.7% higher, and IQ3_S is higher on five of seven domains. ISTA comparisons measure each quant against itsown corresponding BF16 model; they are not direct Swift-versus-Qwen task rankings or proof of equivalent capability.

Thefull result tableincludes the matched Swift starting quant, ISTA comparison and reported error estimates.Evaluation metadatabinds results to the exact released identities. A lexical overlap screen was applied against calibration/development text; it does not establish semantic deduplication or prove absence of overfitting. These 512-token tests do not establish quality at 32K or longer contexts.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#usageUsage

Use a llama.cpp build that supports Qwen3.8. Authenticate with an account granted access while this repository is private.

hf download ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS.gguf --local-dir .

llama-server \
  -m Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS.gguf \
  --jinja -fa on -ngl 99 -c 262144 \
  --temp 1.0 --top-p 0.95 --top-k 20 --min-p 0.0 \
  --presence-penalty 0.0 --repeat-penalty 1.0 --port 8000

Set the context size to fit available memory. The example context setting is not a claim that these quants were evaluated at that length.

This release provides the language-model GGUFs. A Swift 27B vision projector has not been verified for this release, so no projector or validated vision example is included.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#quantization-procedureQuantization procedure

1. Reuse the published ISTA GSQ-RCO per-tensor allocation for each matching Qwen3.8-27B tier.
2. Quantize Swift 1.5 weights with the Swift V1MIX importance matrix and the selected allocation.
3. Apply Swift-specific GSQ refinement. IQ2_XS and IQ3_XXS use the preserved fixed-objective variants; IQ2_S and IQ3_S use their preserved V1MIX variants.
4. Freeze exact file identities, run the reporting evaluations, and preserve the matching MTP packages.

This release reuses ISTA’s allocation search results; it does not claim a new RCO search on Swift. Thearchived recipe recordsdocument construction settings and historical candidates; the release manifest identifies the selected files. The Swift importance matrix is included asimatrix-swift15-v1mix.gguf.

Per-tensor assignments for all eight files are included intensor-allocation. Each dump records the model SHA256, tensor count and type histogram. MTP dumps contain the 851 model tensors plus 15 head tensors.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#methods-and-acknowledgementsMethods and acknowledgements

GSQ and RCO were developed by the Deep Algorithms and Systems Lab at the Institute of Science and Technology Austria. This Swift adaptation is by UkisAI.

• GSQ:paper·code
• RCO:paper·code
• GGUF runtime and conversion:llama.cpp

We acknowledge the Qwen team for the original model, ISTA-DASLab for the quantization methods and published allocations, and NVIDIA Innovation Lab, AWS and Google Cloud for supporting Swift development.

https://huggingface.co/ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF#license-and-accessLicense and access

These adapted weights are distributed under theSwift Open License v1.0. The original Qwen components retain theirApache 2.0 licenseandNOTICE. See the Swift license for applicable terms and enterprise licensing information.