Qwen 3.6 vs Qwen3.8 27B VRAM comparison
If you're evaluating Qwen3.8 27B (Qwen 3.8 27B) against Qwen 3.6 27B for GGUF/Ollama deployment, this table compares total memory under the same formula and context settings.
VRAM totals by quantization and context
| Quantization | Context | Qwen 3.6 27B total | Qwen 3.8 27B total | Delta (3.8 - 3.6) |
|---|---|---|---|---|
| Q5_K_M | 16,384 | 22.55 GiB | 22.55 GiB | +0 GiB |
| Q5_K_M | 32,768 | 26.67 GiB | 26.67 GiB | +0 GiB |
| Q5_K_M | 65,536 | 34.91 GiB | 34.91 GiB | +0 GiB |
| Q4_K_M | 16,384 | 19.88 GiB | 19.88 GiB | +0 GiB |
| Q4_K_M | 32,768 | 24 GiB | 24 GiB | +0 GiB |
| Q4_K_M | 65,536 | 32.24 GiB | 32.24 GiB | +0 GiB |
Continue with model-specific pages
FAQ
Do Qwen 3.6 and 3.8 27B need very different VRAM?
They are close in this baseline because both are 27B-class checkpoints. Practical differences come from runtime implementation, context target, and quantization choice.
Which model should I pick for Ollama first?
Start from the model quality you need, then confirm fit at your target context. If both fit, benchmark your actual prompts before deciding.