Even entry-level 8 GB cards run 4–9B models well. Below are the 29 models we track, with the highest-quality quantization that fits in 8 GB, memory needed and estimated speed (8K context, data updated 2026-09-27).
Calculate with your own settings →
Speed estimated on the RTX 5060 (448 GB/s).
| Model | Best quant | Memory (8K) | Verdict | Speed |
|---|---|---|---|---|
| Qwen3.5 4B Qwen · Lightweight, image input | Q8_0 | 6.0GB | Comfortable | Fast (63 tok/s) |
| Gemma 4 E4B Gemma · Lightweight for phones/laptops | Q6_K | 7.6GB | Tight | Fast (48 tok/s) |
| Qwen3 8B Qwen · The go-to 8B | Q4_K_M | 7.0GB | Tight | Fast (63 tok/s) |
| Llama 3.1 8B Llama · Classic, many fine-tunes | Q5_K_M | 7.6GB | Tight | Fast (55 tok/s) |
| Ministral 3 8B Mistral | Q4_K_M | 7.3GB | Tight | Fast (58 tok/s) |
| DeepSeek-R1-0528-Qwen3-8B DeepSeek · Reasoning (thinking) model | Q4_K_M | 7.0GB | Tight | Fast (63 tok/s) |
| llm-jp-4 8B thinking Japanese · Japanese-focused (Japan) | Q4_K_M | 7.1GB | Tight | Fast (60 tok/s) |
| Qwen3-Swallow 8B Japanese · Japanese-enhanced (Institute of Science Tokyo) | Q4_K_M | 7.0GB | Tight | Fast (63 tok/s) |
| Qwen3.5 9B Qwen · Image input | Q4_K_M | 6.9GB | Comfortable | Fast (54 tok/s) |
| Gemma 4 12B Gemma · Mid-size, good multilingual | Q3_K_M | 7.2GB | Tight | Fast (53 tok/s) |
| Model | Best quant | Memory (8K) | Verdict | Speed |
|---|---|---|---|---|
| Ministral 3 14B Mistral | Q4_K_M | 10.7GB | Spills to RAM | Slowish (9.8 tok/s) |
| Qwen3 14B Qwen | Q4_K_M | 11.2GB | Spills to RAM | Slowish (8.6 tok/s) |
| gpt-oss-20b OpenAI · OpenAI open-weight model (MXFP4) | Q4_K_M (MXFP4) | 13.2GB | Spills to RAM | Usable (20 tok/s) |
| Devstral Small 2 24B Mistral · Coding | Q4_K_M | 16.9GB | Spills to RAM | Very slow (3.6 tok/s) |
| Gemma 4 26B-A4B Gemma · Fast MoE | Q4_K_M | 17.0GB | Spills to RAM | Slowish (14 tok/s) |
| Qwen3.8 27B Qwen · Latest generation (Aug 2026) | Q4_K_M | 18.4GB | Spills to RAM | Very slow (3.1 tok/s) |
| Qwen3-Coder 30B-A3B Qwen · Coding MoE | Q4_K_M | 20.4GB | Spills to RAM | Slowish (15 tok/s) |
| GLM-4.7-Flash GLM · MoE, strong at long context | Q4_K_M | 20.4GB | Spills to RAM | Usable (16 tok/s) |
| Nemotron 3 Nano 30B-A3B NVIDIA · Mamba hybrid, light long context | Q4_K_M | 20.3GB | Spills to RAM | Slowish (14 tok/s) |
| Gemma 4 31B Gemma · Largest Gemma 4 | Q4_K_M | 21.5GB | Spills to RAM | Very slow (2.4 tok/s) |
| Qwen3 32B Qwen | Q4_K_M | 23.1GB | Spills to RAM | Very slow (2.2 tok/s) |
| Qwen3-Swallow 32B Japanese · Japanese-enhanced (Institute of Science Tokyo) | Q4_K_M | 23.1GB | Spills to RAM | Very slow (2.2 tok/s) |
| llm-jp-4 33B thinking Japanese · Japanese-focused (Japan) | Q4_K_M | 23.4GB | Spills to RAM | Very slow (2.2 tok/s) |
| Qwen3.6 35B-A3B Qwen · Popular fast MoE | Q4_K_M | 23.1GB | Spills to RAM | Usable (16 tok/s) |