RunLLMLocal
Open the checker

Compare · Models

Mistral Small 3 24B vs Gemma 3 27B: which runs better locally?

Mistral Small 3 24B needs about 17.7 GB at Q4 with 8K context and fits fully on 5 of 40 GPUs. Gemma 3 27B needs 22.4 GB and fits on 1.

VerdictMistral Small 3 24B is easier to run

Mistral Small 3 24B needs about 17.7 GB at Q4 with 8K context and fits fully on 5 of 40 GPUs. Gemma 3 27B needs 22.4 GB and fits on 1.

Mistral Small 3 24BChat, Vision
Parameters24B
Memory at Q4, 8K17.7 GB
Memory at Q8, 8K30.1 GB
GPUs that run it fully5 / 40
Cheapest desktop cardRX 7900 XT (20 GB)
Gemma 3 27BChat, Vision
Parameters27.4B
Memory at Q4, 8K22.4 GB
Memory at Q8, 8K36.5 GB
GPUs that run it fully1 / 40
Cheapest desktop cardRTX 5090 (32 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.

GPUMistral Small 3 24BGemma 3 27B
RTX 4090 24 GBRuns~39 tok/sQ5_K_MOffload~5.9 tok/sQ4_K_M
RX 7900 XTX 24 GBRuns~37 tok/sQ5_K_MOffload~5.7 tok/sQ4_K_M
RTX 3090 / 3090 Ti 24 GBRuns~36 tok/sQ5_K_MOffload~5.5 tok/sQ4_K_M
RX 7900 XT 20 GBRuns~37 tok/sQ4_K_MOffload~4.3 tok/sQ4_K_M
RTX 5090 32 GBRuns~59 tok/sQ6_KRuns~60 tok/sQ5_K_M
RTX 5080 16 GBOffload~5.8 tok/sQ4_K_MOffload~4.5 tok/sQ4_K_M
RTX 5070 Ti 16 GBOffload~5.5 tok/sQ4_K_MOffload~4.3 tok/sQ4_K_M
RTX 4080 / 4080 Super 16 GBOffload~4.6 tok/sQ4_K_MOffload~3.6 tok/sQ4_K_M
RTX 4070 Ti Super 16 GBOffload~4.2 tok/sQ4_K_MOffload~3.4 tok/sQ4_K_M
RX 9070 XT 16 GBOffload~4.0 tok/sQ4_K_MOffload~3.3 tok/sQ4_K_M
RX 7800 XT 16 GBOffload~4.0 tok/sQ4_K_MOffload~3.2 tok/sQ4_K_M
Arc A770 16 GBOffload~3.6 tok/sQ4_K_MOffload~2.9 tok/sQ4_K_M
RX 6800 XT 16 GBOffload~3.3 tok/sQ4_K_MOffload~2.8 tok/sQ4_K_M
RTX 5060 Ti 16 GBOffload~3.0 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RTX 4060 Ti 16 GBOffload~2.1 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M
RTX 5070 12 GBOffload~3.8 tok/sQ4_K_MOffload~3.1 tok/sQ4_K_M
RTX 4070 / 4070 Super 12 GBOffload~3.1 tok/sQ4_K_MOffload~2.6 tok/sQ4_K_M
Arc B580 12 GBOffload~2.9 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RX 7700 XT 12 GBOffload~2.8 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RX 6700 XT 12 GBOffload~2.6 tok/sQ4_K_MOffload~2.2 tok/sQ4_K_M
RTX 3060 12 GBOffload~2.5 tok/sQ4_K_MOffload~2.2 tok/sQ4_K_M
GTX 1080 Ti 11 GBOffload~3.0 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RTX 3080 10 GBOffload~3.9 tok/sQ4_K_MOffload~3.1 tok/sQ4_K_M
Arc A750 8 GBOffload~2.9 tok/sQ4_K_MOffload~2.5 tok/sQ4_K_M
RTX 5060 Ti 8 GBOffload~2.7 tok/sQ4_K_MOffload~2.3 tok/sQ4_K_M
RTX 5060 8 GBOffload~2.7 tok/sQ4_K_MOffload~2.3 tok/sQ4_K_M
RTX 3070 / 3070 Ti 8 GBOffload~2.7 tok/sQ4_K_MOffload~2.3 tok/sQ4_K_M
RTX 3060 Ti 8 GBOffload~2.7 tok/sQ4_K_MOffload~2.3 tok/sQ4_K_M
RTX 4060 Ti 8 GBOffload~2.3 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RX 7600 8 GBOffload~2.3 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 4060 8 GBOffload~2.2 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 3050 8 GBOffload~2.1 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M
GTX 1660 / 1660 Super 6 GBOffload~2.4 tok/sQ4_K_MOffload~2.1 tok/sQ4_K_M
GTX 1650 4 GBOffload~2.1 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M
RTX 4090 Laptop 16 GBOffload~3.7 tok/sQ4_K_MOffload~3.0 tok/sQ4_K_M
RTX 4080 Laptop 12 GBOffload~2.8 tok/sQ4_K_MOffload~2.4 tok/sQ4_K_M
RTX 4070 Laptop 8 GBOffload~2.2 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 4060 Laptop 8 GBOffload~2.2 tok/sQ4_K_MOffload~2.0 tok/sQ4_K_M
RTX 3060 Laptop 6 GBOffload~2.4 tok/sQ4_K_MOffload~2.1 tok/sQ4_K_M
RTX 4050 Laptop 6 GBOffload~2.1 tok/sQ4_K_MOffload~1.9 tok/sQ4_K_M

Which should you run?

Mistral Small 3 24B is easier to run. Mistral Small 3 24B needs about 17.7 GB at Q4 with 8K context and fits fully on 5 of 40 GPUs. Gemma 3 27B needs 22.4 GB and fits on 1.

Only the Mistral Small 3 24B fits on RTX 4090, RTX 3090 / 3090 Ti, RX 7900 XTX and RX 7900 XT.

See every result on its own page: Mistral Small 3 24B and Gemma 3 27B.

Questions

Is Mistral Small 3 24B or Gemma 3 27B better?

Mistral Small 3 24B is easier to run. Mistral Small 3 24B needs about 17.7 GB at Q4 with 8K context and fits fully on 5 of 40 GPUs. Gemma 3 27B needs 22.4 GB and fits on 1. The right pick depends on what you use it for: compare the strengths listed on each model's page.

Which needs less VRAM, Mistral Small 3 24B or Gemma 3 27B?

Mistral Small 3 24B needs about 18 GB at Q4 with 8K context, Gemma 3 27B about 22 GB. At Q8 it is 30 GB versus 37 GB.

Can I run Mistral Small 3 24B and Gemma 3 27B on a 12 GB GPU?

On an RTX 3060 12 GB, Mistral Small 3 24B only runs with part of the model in system RAM, at a few tokens per second and Gemma 3 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