How much memory does Qwen3.8 27B need?
About 19 GB at Q4_K_M with an 8K context. It runs on a Mac with 32 GB of unified memory or more.
Try other settings in the calculatorMemory by quantization and context
Total memory: weights, KV cache in F16 and about 1 GB of buffers. Under each total, the smallest Mac that can run it with the default GPU memory limit.
| Quantization | 8K context | 32K context | 128K context |
|---|---|---|---|
| Q3_K_M | 15 GB24 GB Mac | 17 GB24 GB Mac | 23 GB32 GB Mac |
| Q4_K_M | 19 GB32 GB Mac | 20 GB32 GB Mac | 27 GB48 GB Mac |
| Q5_K_M | 21 GB32 GB Mac | 23 GB32 GB Mac | 29 GB48 GB Mac |
| Q6_K | 24 GB48 GB Mac | 26 GB48 GB Mac | 32 GB48 GB Mac |
| Q8_0 | 31 GB48 GB Mac | 33 GB48 GB Mac | 39 GB64 GB Mac |
| F16 | 57 GB96 GB Mac | 59 GB96 GB Mac | 65 GB96 GB Mac |
About the model
- Parameters
- 27.78B
- Type
- Dense
- Attention
- hybrid: linear attention plus full attention
- KV cache, F16
- 0.5 GB at 8K, 8.6 GB at 128K
- On Hugging Face
- Qwen/Qwen3.8-27B
Most layers use linear attention, which keeps a small fixed-size state instead of a cache. Only the full-attention layers grow with context, so long contexts are cheap.
Questions
How much memory does Qwen3.8 27B need?
About 19 GB at Q4_K_M with an 8K context: 17 GB of weights, 0.5 GB of KV cache and about 1 GB of buffers. At 32K context it needs about 20 GB.
Can Qwen3.8 27B run on a Mac?
Yes. At Q4_K_M with an 8K context it fits on a Mac with 32 GB of unified memory or more, using the default GPU memory limit of about three quarters.
How big is the Qwen3.8 27B KV cache?
About 0.5 GB at 8K tokens and 8.6 GB at 128K tokens with an F16 cache. Its attention design: hybrid: linear attention plus full attention.
Other models
- Gemma 4 12B
- gpt-oss 20B
- Gemma 4 26B-A4B
- GLM-4.7 Flash
- Gemma 4 31B
- Qwen3.6 35B-A3B
- gpt-oss 120B
- Mistral Medium 3.5
- Qwen3.8 Flash Next
- DeepSeek V4 Flash
- GLM-5.3
- DeepSeek V4 Pro
- Kimi K3
Data read from Hugging Face on 2026-11-09. These are estimates; benchmarks on this site report measured memory. How the estimate works. Compare all models in the model fit chart.
Get the next one
New benchmarks and agent runs by email, with the raw data. Or follow on RSS or X.