Can I run it? / PC without a GPU
Best LLMs to run on a PC or laptop without a GPU
Updated November 2026
As of November 2026, the best model to run on a PC or laptop without a GPU is Ternary Bonsai 2 27B by PrismML (Small LM Index 71.2), at about 5 tokens/s with Q2_0 (as released) weights. For faster replies, LFM3 1.7B (61.2) runs at about 40 tokens/s. On the 16 GB DDR4 laptop, Nanbeige4.2 3B (67.2) fits.
Without a graphics card, models run from system RAM at about 50–90 GB/s. Small dense models and Mixture-of-Experts models with few active parameters run at usable speeds. About 27 GB is usable for a model, at 83 GB/s. The figures below assume 4-bit weights and a 8K-token context, and only include models that generate at least 5 tokens per second.
Models that run well on a PC or laptop without a GPU
The 25 best of the 73 models that run at 5+ tokens/s, ranked by the Small LM Index.
| # | Model | Index | Memory | Speed |
|---|---|---|---|---|
| 1 | Ternary Bonsai 2 27B PrismML · Q2_0 (as released) · Apache-2.0 | 71.2 | 9.3 GB | ≈ 5 tok/s |
| 2 | Qwen3.6 35B-A3B Alibaba (Qwen) · 4-bit · Apache-2.0 | 68.3 | 23 GB | ≈ 11 tok/s |
| 3 | Nanbeige4.2 3B Nanbeige (Kanzhun) · 4-bit · Apache-2.0 | 67.2 | 3.8 GB | ≈ 16 tok/s |
| 4 | Qwen3.5 35B-A3B Alibaba (Qwen) · 4-bit · Apache-2.0 | 66.9 | 23 GB | ≈ 11 tok/s |
| 5 | Spark-X2.6 4B SparkLLM (XHToken) · 4-bit · Apache-2.0 | 66.5 | 3.4 GB | ≈ 17 tok/s |
| 6 | GLM-4.7-Flash Z.ai (GLM) · 4-bit · MIT | 66.0 | 20 GB | ≈ 11 tok/s |
| 7 | Spark-X2.5 4B SparkLLM (XHToken) · 4-bit · Apache-2.0 | 65.9 | 3.4 GB | ≈ 17 tok/s |
| 8 | Qwen3 30B-A3B Thinking 2507 Alibaba (Qwen) · 4-bit · Apache-2.0 | 65.4 | 20 GB | ≈ 10 tok/s |
| 9 | Gemma 4 26B-A4B Google · 4-bit · Apache-2.0 | 64.8 | 17 GB | ≈ 9 tok/s |
| 10 | North Mini Code Cohere · 4-bit · Unverified | 64.7 | 21 GB | ≈ 11 tok/s |
| 11 | gpt-oss-20b OpenAI · 4-bit · Apache-2.0 | 64.7 | 14 GB | ≈ 10 tok/s |
| 12 | Nemotron 3 Nano 30B-A3B NVIDIA · 4-bit · NVIDIA Open Model | 64.4 | 20 GB | ≈ 10 tok/s |
| 13 | Qwen3.5 9B Alibaba (Qwen) · 4-bit · Apache-2.0 | 64.1 | 6.4 GB | ≈ 8 tok/s |
| 14 | Nemotron 3.5 Lightning NVIDIA · 4-bit · OpenMDW-1.1 | 63.3 | 22 GB | ≈ 10 tok/s |
| 15 | LFM2.5 24B-A2B Liquid AI · 4-bit · LFM Open License | 63.1 | 16 GB | ≈ 17 tok/s |
| 16 | Edge0 35B-A3B (preview) Edge0 · 4-bit · Apache-2.0 | 63.0 | 22 GB | ≈ 11 tok/s |
| 17 | Sarvam 30B Sarvam AI · 4-bit · Apache-2.0 | 62.8 | 21 GB | ≈ 14 tok/s |
| 18 | Granite 4.2 8B IBM (Granite) · 4-bit · Apache-2.0 | 62.7 | 6.9 GB | ≈ 9 tok/s |
| 19 | LFM3 1.7B Liquid AI · 4-bit · LFM Open License | 61.2 | 1.7 GB | ≈ 40 tok/s |
| 20 | MiniCPM5 2B OpenBMB (MiniCPM) · 4-bit · Apache-2.0 | 60.9 | 2.4 GB | ≈ 27 tok/s |
| 21 | Mellum2.1 12B-A2.5B Thinking JetBrains (Mellum) · 4-bit · Apache-2.0 | 60.8 | 8.3 GB | ≈ 14 tok/s |
| 22 | Qwen3 4B Thinking 2507 Alibaba (Qwen) · 4-bit · Apache-2.0 | 59.8 | 4.2 GB | ≈ 17 tok/s |
| 23 | Qwen3.5 4B Alibaba (Qwen) · 4-bit · Apache-2.0 | 59.7 | 3.3 GB | ≈ 17 tok/s |
| 24 | Nemotron Labs Dory 1.0 4B NVIDIA · 4-bit · OpenMDW-1.1 | 58.6 | 3.4 GB | ≈ 17 tok/s |
| 25 | Falcon H1R 7B TII (Falcon) · 4-bit · Falcon LLM License | 58.6 | 5.6 GB | ≈ 9 tok/s |
Change the context length or quantization, or see models that only just fit, in Can I run it? for this device.
How to run Ternary Bonsai 2 27B on a PC or laptop without a GPU
- Easiest: Ollama or LM Studio. Both run on the CPU when there is no supported GPU.
- Make it faster. Use dual-channel memory (two RAM sticks), close other apps, and prefer Mixture-of-Experts models with few active parameters.
Every model page lists the same commands for its own files. Ternary Bonsai 2 27B's page also has a memory table for other context lengths.
Questions
What is the best LLM for a PC or laptop without a GPU?
As of November 2026, the best model to run on a PC or laptop without a GPU is Ternary Bonsai 2 27B by PrismML (Small LM Index 71.2), at about 5 tokens/s with Q2_0 (as released) weights. For faster replies, LFM3 1.7B (61.2) runs at about 40 tokens/s. On the 16 GB DDR4 laptop, Nanbeige4.2 3B (67.2) fits.
What is the biggest model that runs on a PC or laptop without a GPU?
MiniCPM-V 4.7 35B-A3B (35.2B parameters, 3B active) fits at 4-bit, needing 23 GB and generating about 11 tokens/s. Bigger models may fit at lower quality or run more slowly; try them in the Can I run it? tool.
How much memory can a model use on a PC or laptop without a GPU?
About 27 GB of the 32 GB. Without a graphics card, models run from system RAM at about 50–90 GB/s. Small dense models and Mixture-of-Experts models with few active parameters run at usable speeds.
How accurate are these speeds?
They are estimates from memory bandwidth (83 GB/s here), which limits how fast a model generates text, with a typical efficiency for this kind of hardware. Expect real speeds within about a third either way. Long prompts take longer to read before the first word appears.