Qwen · Apache 2.0
Dense 27B from the Qwen 3.6 family (April 2026). Strong coding and agentic workflows; native multimodal. 262K context window. VRAM figures include modest KV headroom at default context; long-context runs need more memory or KV cache quantization. Apache 2.0.
Parameters
27B
VRAM
20 GB
Context
256K
Formats
4
GPUs
41
Qwen3.6-27B (27B) requires 20 GB VRAM at recommended quality (Q5_K_M). At efficient quality (Q4_K_M), it fits in 17 GB VRAM, making it compatible with the NVIDIA RTX 4090 Laptop (150-175W). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 150 tok/s at Q8_0. For the best experience, AMD AI Powerhouse ($1,818) is recommended.
Source: OwnRig methodology
20 GB
Q5_K_M
19 GB
256K tokens
Chat
| Quality | Quantization | VRAM | File Size |
|---|---|---|---|
| full | Q8_0 | 30 GB | 28 GB |
| recommended | Q5_K_M | 20 GB | 19 GB |
| efficient | Q4_K_M | 17 GB | 16 GB |
| compressed | Q3_K_M | 15 GB | 14 GB |
KV cache VRAM at Q5_K_M quality. Longer context = more memory.
| Context | KV Cache | Total VRAM |
|---|---|---|
| 2K | 205 MB | 20.2 GB |
| 4K | 410 MB | 20.4 GB |
| 8K | 819 MB | 20.8 GB |
| 16K | 1.7 GB | 21.7 GB |
| 32K | 3.4 GB | 23.4 GB |
| 64K | 6.7 GB | 26.7 GBexceeds 24 GB |
| 128K | 13.4 GB | 33.4 GBexceeds 24 GB |
Showing 41 of 41 entries
Qwen3.6-27B is commonly used with Continue, LM Studio, Open WebUI. For an AI coding workflow, pair it with an embedding model like nomic-embed-text for local RAG.
Data confidence: community. Source
VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Qwen is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.