Qwen3.8-27B IQ4_XS 量化模型（适配 16GB 显存）
本模型基于 Qwen3.8-27B 进行 IQ4_XS 量化（4‑bit），文件体积为 12.9 GiB，专为 16GB 显存 的显卡优化，在保持较低困惑度的同时，兼顾推理速度和显存占用。
与同体积的 UD_IQ3_K_XL（12.5 GiB）量化方案进行了全面对比，评估指标如下。
📊 量化质量对比
| 评估指标 | IQ4_XS (本模型) | UD_IQ3_K_XL (对比) |
|---|---|---|
| 文件大小 | 12.9 GB | 12.5 GB |
| 量化精度 | IQ4_XS (4‑bit) | UD_IQ3_K_XL (约 3‑bit?) |
| 量化模型困惑度 (Mean PPL) | 7.0875 ± 0.0459 | 7.1117 ± 0.0459 |
| 与基座模型 PPL 相关性 | 99.28% | 99.31% |
| 平均 KL 散度 (Mean KLD) | 0.03168 ± 0.00029 | 0.03130 ± 0.00032 |
| 最大 KL 散度 (Max KLD) | 12.089（更小） | 21.409 |
| 99.9% KL 分位数 | 1.134 | 1.219 |
| Top‑1 一致率 (Same top p) | 91.845% ± 0.071% | 92.419% ± 0.069% |
| 平均概率变化 (Mean Δp) | -0.381% ± 0.013%（更接近 0） | -0.738% ± 0.013% |
| RMS 概率变化 (RMS Δp) | 4.893% ± 0.041%（更小） | 5.120% ± 0.046% |
在不启用MTP的情况下可以做到16Gib净空VRAM（不作为Windows的显示显卡）的情况下110k上下文 开启MTP大概80k上下文
Qwen3.8-27B IQ4_XS Quantized Model (Optimized for 16GB VRAM)
This model is a IQ4_XS quantization (4‑bit) of Qwen3.8-27B. The file size is 12.9 GB, and it is specifically optimized for graphics cards with 16GB VRAM, offering a good balance between low perplexity, inference speed, and memory usage.
We have conducted a comprehensive comparison with the UD_IQ3_K_XL quantization scheme (12.5 GB) of similar size. The evaluation metrics are shown below.
📊 Quantization Quality Comparison
| Metric | IQ4_XS (This Model) | UD_IQ3_K_XL (Comparison) |
|---|---|---|
| File Size | 12.9 GB | 12.5 GB |
| Quantization Precision | IQ4_XS (4‑bit) | UD_IQ3_K_XL (~3‑bit?) |
| Quantized Model Perplexity (Mean PPL) | 7.0875 ± 0.0459 | 7.1117 ± 0.0459 |
| Correlation with Base PPL | 99.28% | 99.31% |
| Mean KL Divergence (Mean KLD) | 0.03168 ± 0.00029 | 0.03130 ± 0.00032 |
| Maximum KL Divergence (Max KLD) | 12.089 (smaller) | 21.409 |
| 99.9% KLD Quantile | 1.134 | 1.219 |
| Top‑1 Agreement (Same top p) | 91.845% ± 0.071% | 92.419% ± 0.069% |
| Mean Probability Change (Mean Δp) | -0.381% ± 0.013% (closer to 0) | -0.738% ± 0.013% |
| RMS Probability Change (RMS Δp) | 4.893% ± 0.041% (smaller) | 5.120% ± 0.046% |
Context Length with 16GB VRAM
- Without MTP (Multi‑Token Prediction): Can achieve a context length of 110k tokens when there is 16GB free VRAM (and the GPU is not used as the primary display adapter for Windows).
- With MTP enabled: Approximately 80k tokens.