Swift-1.5-Qwen3.8-27B-heretic-GSQ-RCO-GGUF
This is a partially decensored build of UkisAI's Swift 1.5 Qwen3.8-27B, made with Heretic. It is packaged as a single-file llama.cpp GGUF and quantized the same way as UkisAI's own GSQ-RCO -mtp releases, using the same per-tensor allocations and the same imatrix. Three sizes are provided: IQ3S, IQ3XXS and IQ2S.
Params
27.0 B
Context
—
Downloads 30d
1.1 M
Likes
33
Download history
tracking started — chart appears after 7 days of snapshots (6 recorded)1.1 M downloads in the last 30 days
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| Swift-1.5-Qwen3.8-27B-heretic-mmproj-F16.gguf | GGUF | 0.9 GB | 5.6 GB | ✅ Runs comfortably |
| Swift-1.5-Qwen3.8-27B-heretic-GSQ-RCO-IQ2_S-mtp.gguf | IQ2_S | 9.6 GB | 15.1 GB | ✅ Runs comfortably |
| Swift-1.5-Qwen3.8-27B-heretic-GSQ-RCO-IQ3_XXS-mtp.gguf | IQ3_XXS | 10.4 GB | 16.0 GB | ✅ Runs comfortably |
| Swift-1.5-Qwen3.8-27B-heretic-GSQ-RCO-IQ3_S-mtp.gguf | IQ3_S | 12.1 GB | 17.9 GB | ✅ Runs comfortably |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Specifications
- Parameters
- 27.0 B
- Licence
- other
- First seen on the Hub
- 2026-09-29
- Training datasets
- undisclosed
- IFEval (reported)
- 96.52
- GSM8K (250-question random sample) (reported)
- 97.6
- MMLU-Pro (300-question random sample) (reported)
- 86.67
- Added to our catalog
- 2026-11-05
Compare with any image-text-to-text model