Can I run Qwen 3.6 27B on an RTX 5090 (32GB)?
Computed with the open FitLLM engine — accurate per-layer KV-cache modeling, not a naive estimate. Updated 2026-10-16.
Memory breakdown (Q4_K_M, F16 KV, 33K context)
| Model weights | 15.5 GB |
| KV cache | 2.0 GB |
| Linear-attention state (fixed — does not grow with context) | 0.1 GB |
| Runtime overhead + reserve | 5.1 GB |
| Total used | 22.8 / 32 GB |
| Free | 9.2 GB |
Max context that fits at Q4_K_M: ~109K tokens · with Q8 KV cache → ~167K tokens.
Every quantization on the RTX 5090
| Weight quant | Weights | Fits (KV F16) | Used @32K |
|---|---|---|---|
| Q4_K_M | 15.5 GB | ✅ up to 109K ctx | 22.8 / 32.0 GB |
| Q5_K_M | 18.1 GB | ✅ up to 80K ctx | 25.7 / 32.0 GB |
| Q6_K | 20.8 GB | ✅ up to 50K ctx | 28.7 / 32.0 GB |
| Q8_0 | 26.9 GB | ❌ won't fit | 35.6 / 32.0 GB |
| FP16 | 50.7 GB | ❌ won't fit | 62.2 / 32.0 GB |
Lower weight quants free memory at some output-quality cost — Q4 is the common sweet spot; below that quality drops faster.
KV cache is F16 here (llama.cpp default). Drop it to Q8/Q4 (-ctk/-ctv) for more context.
Embed this verdict
Live badge for your README or model card — recomputed by the engine, never stale:
[](https://fitllm.run/can-i-run/qwen-3-6-27b-on-rtx-5090)
Or from your terminal (exit 0/1 — works as a pre-download guard):
npx fitllm-engine "Qwen 3.6 27B" --gpu "RTX 5090"Why most VRAM calculators get this wrong
Qwen 3.6 27B is a hybrid model: only 16 of its 64 layers use full attention — the rest are linear and keep no growing KV cache. Naive calculators count every layer at full context and badly over-estimate.
Other options
same GPU Models that fit on the RTX 5090: GLM-4.7-Flash, gpt-oss-20b, Qwen 3.6 27B, Qwen 3.6 35B-A3B, Qwen-AgentWorld-35B-A3B, Gemma 4 e2b, Gemma 4 e4b, Gemma 4 12b, Gemma 4 26b A4B, Gemma 4 31b, Llama-3.2-3B-Instruct, Llama-3.1-8B-Instruct, MiniCPM5-1B, Qwen3-0.6B, Qwen3-1.7B, Llama-3.2-1B-Instruct, Gemma-3-1B-it, Qwen 3.8 27B, Laguna XS 2.1, Spark-X2.5-4B, Granite-4.2-30B, MiniCPM5-2B, Nex-N2.5-mini.
same model GPUs that run Qwen 3.6 27B: RTX 5090 (32GB), RTX 4090 (24GB), RTX 3090 (24GB), RTX 3090 Ti (24GB), RTX 6000 Ada (48GB), RTX PRO 6000 Blackwell (96GB), RX 7900 XTX (24GB), Radeon PRO W7900 (48GB), 2× RTX 3090 (48GB), 2× RTX 4090 (48GB), 4× RTX 3090 (96GB), A100 40GB (40GB), A100 80GB (80GB), H100 80GB (80GB), H200 141GB (141GB), B200 (180GB), RTX PRO 5000 Blackwell 48GB (48GB), RTX PRO 5000 Blackwell 72GB (72GB), RTX PRO 4500 Blackwell (32GB), RTX PRO 4000 Blackwell (24GB), DGX Spark 128GB (128GB), 2× DGX Spark 128GB (256GB), 4× DGX Spark 128GB (512GB).
Reproduce it
Qwen 3.6 27B = 27.2B, 64 layers. The RTX 5090 has 32GB / 1792GB/s. Same math, open source: fitllm-engine. GGUF bpw from llama.cpp.
All numbers are computed by the open-source fitllm-engine (MIT) from official model config.json values — reproduce or audit them yourself. Estimates; real usage varies with runtime (llama.cpp / MLX / Ollama), driver and display. Found a mismatch? Report it. · FitLLM home