RunLLMLocal
Open the checker

Compare · Models

Gemma 4 31B vs Qwen3.6 27B: which runs better locally?

Gemma 4 31B needs about 24.6 GB at Q4 with 8K context and fits fully on 1 of 40 GPUs. Qwen3.6 27B needs 19.0 GB and fits on 4.

VerdictQwen3.6 27B is easier to run

Gemma 4 31B needs about 24.6 GB at Q4 with 8K context and fits fully on 1 of 40 GPUs. Qwen3.6 27B needs 19.0 GB and fits on 4.

Gemma 4 31BChat, Vision, Reasoning
Parameters31B
Memory at Q4, 8K24.6 GB
Memory at Q8, 8K40.7 GB
GPUs that run it fully1 / 40
Cheapest desktop cardRTX 5090 (32 GB)
Qwen3.6 27BChat, Coding, Reasoning, Vision
Parameters27.5B
Memory at Q4, 8K19.0 GB
Memory at Q8, 8K33.2 GB
GPUs that run it fully4 / 40
Cheapest desktop cardRTX 3090 / 3090 Ti (24 GB)

Each GPU at 8K context with the best quantization that fits. Offload assumes 32 GB of system RAM. Green marks the better result. Rows where the two differ come first.

GPUGemma 4 31BQwen3.6 27B
RTX 4090 24 GBOffload~4.9 tok/sQ4_K_MRuns~40 tok/sQ4_K_M
RX 7900 XTX 24 GBOffload~4.7 tok/sQ4_K_MRuns~39 tok/sQ4_K_M
RTX 3090 / 3090 Ti 24 GBOffload~4.6 tok/sQ4_K_MRuns~38 tok/sQ4_K_M
RTX 5090 32 GBRuns~53 tok/sQ5_K_MRuns~52 tok/sQ6_K
RX 7900 XT 20 GBOffload~3.6 tok/sQ4_K_MOffload~4.7 tok/sQ4_K_M
RTX 5080 16 GBOffload~3.8 tok/sQ4_K_MOffload~4.9 tok/sQ4_K_M
RTX 5070 Ti 16 GBOffload~3.6 tok/sQ4_K_MOffload~4.6 tok/sQ4_K_M
RTX 4080 / 4080 Super 16 GBOffload~3.1 tok/sQ4_K_MOffload~3.9 tok/sQ4_K_M
RTX 4070 Ti Super 16 GBOffload~2.9 tok/sQ4_K_MOffload~3.6 tok/sQ4_K_M
RX 9070 XT 16 GBOffload~2.8 tok/sQ4_K_MOffload~3.4 tok/sQ4_K_M
RX 7800 XT 16 GBOffload~2.7 tok/sQ4_K_MOffload~3.4 tok/sQ4_K_M
Arc A770 16 GBOffload~2.5 tok/sQ4_K_MOffload~3.1 tok/sQ4_K_M
RX 6800 XT 16 GBOffload~2.4 tok/sQ4_K_MOffload~2.9 tok/sQ4_K_M
RTX 5060 Ti 16 GBOffload~2.2 tok/sQ4_K_MOffload~2.6 tok/sQ4_K_M
RTX 4060 Ti 16 GBOffload~1.7 tok/sQ4_K_MOffload~1.8 tok/sQ4_K_M
RTX 5070 12 GBOffload~2.6 tok/sQ4_K_MOffload~3.2 tok/sQ4_K_M
RTX 4070 / 4070 Super 12 GBOffload~2.3 tok/sQ4_K_MOffload~2.7 tok/sQ4_K_M
Arc B580 12 GBOffload~2.1 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RX 7700 XT 12 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RX 6700 XT 12 GBOffload~2.0 tok/sQ4_K_MOffload~2.2 tok/sQ4_K_M
RTX 3060 12 GBOffload~1.9 tok/sQ4_K_MOffload~2.2 tok/sQ4_K_M
GTX 1080 Ti 11 GBOffload~2.2 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RTX 3080 10 GBOffload~2.7 tok/sQ4_K_MOffload~3.3 tok/sQ4_K_M
Arc A750 8 GBOffload~2.2 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RTX 5060 Ti 8 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 5060 8 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 3070 / 3070 Ti 8 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 3060 Ti 8 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 4060 Ti 8 GBOffload~1.8 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RX 7600 8 GBOffload~1.8 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 4060 8 GBOffload~1.8 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 3050 8 GBOffload~1.7 tok/sQ4_K_MOffload~1.8 tok/sQ4_K_M
GTX 1660 / 1660 Super 6 GBOffload~2.3 tok/sQ3_K_MOffload~2.1 tok/sQ4_K_M
GTX 1650 4 GBOffload~2.1 tok/sQ3_K_MOffload~1.9 tok/sQ4_K_M
RTX 4090 Laptop 16 GBOffload~2.6 tok/sQ4_K_MOffload~3.1 tok/sQ4_K_M
RTX 4080 Laptop 12 GBOffload~2.1 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 4070 Laptop 8 GBOffload~1.8 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M
RTX 4060 Laptop 8 GBOffload~1.8 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M
RTX 3060 Laptop 6 GBOffload~2.3 tok/sQ3_K_MOffload~2.1 tok/sQ4_K_M
RTX 4050 Laptop 6 GBOffload~2.1 tok/sQ3_K_MOffload~1.9 tok/sQ4_K_M

Which should you run?

Qwen3.6 27B is easier to run. Gemma 4 31B needs about 24.6 GB at Q4 with 8K context and fits fully on 1 of 40 GPUs. Qwen3.6 27B needs 19.0 GB and fits on 4.

Only Qwen3.6 27B fits on RTX 4090, RTX 3090 / 3090 Ti and RX 7900 XTX.

See every result on its own page: Gemma 4 31B and Qwen3.6 27B.

Questions

Is Gemma 4 31B or Qwen3.6 27B better?

Qwen3.6 27B is easier to run. Gemma 4 31B needs about 24.6 GB at Q4 with 8K context and fits fully on 1 of 40 GPUs. Qwen3.6 27B needs 19.0 GB and fits on 4. The right pick depends on what you use it for: compare the strengths listed on each model's page.

Which needs less VRAM, Gemma 4 31B or Qwen3.6 27B?

Gemma 4 31B needs about 25 GB at Q4 with 8K context, Qwen3.6 27B about 19 GB. At Q8 it is 41 GB versus 33 GB.

Can I run Gemma 4 31B and Qwen3.6 27B on a 12 GB GPU?

On an RTX 3060 12 GB, Gemma 4 31B only runs with part of the model in system RAM, at a few tokens per second and Qwen3.6 27B only runs with part of the model in system RAM, at a few tokens per second.

More comparisons

Gemma 3 27B vs Gemma 4 31BModel comparisonQwen3 14B vs Gemma 4 12BModel comparisonQwen3 8B vs Llama 3.1 8BModel comparisongpt-oss 20B vs Qwen3 30B-A3B (MoE)Model comparisonQwen3 30B-A3B (MoE) vs Qwen3 32BModel comparisonQwen3.6 35B-A3B (MoE) vs Qwen3 30B-A3B (MoE)Model comparison

Estimates from memory size and bandwidth at 8K context. Real speed varies with software, drivers and settings.

Copied