Best small open language models
Small language models with 4 billion parameters or fewer can run on modest hardware, including laptops and phones. They answer quickly, and they are a common base for fine-tuning on a narrow task.
Last updated 3 November 2026 · refreshed after every Hugging Face sync · 15 models
| # | Model | Size | Memory | License | Downloads (30 days) | Likes |
|---|---|---|---|---|---|---|
| 1 | Qwen3-0.6B Qwen | 600M · FP16 | ~2 GB | Apache 2.0 | 23.2M | 1.9K |
| 2 | Qwen3-4B Qwen | 4B · FP16 | ~10 GB | Apache 2.0 | 8.4M | 780 |
| 3 | Qwen2.5-1.5B-Instruct Qwen | 1.5B · FP16 | ~4 GB | Apache 2.0 | 4.6M | 926 |
| 4 | Qwen2.5-0.5B-Instruct Qwen | 500M · FP16 | ~2 GB | Apache 2.0 | 4.9M | 726 |
| 5 | Qwen3-1.7B Qwen | 1.7B · FP16 | ~5 GB | Apache 2.0 | 4.2M | 611 |
| 6 | Qwen3-4B-Instruct-2507 Qwen | 4B · FP16 | ~10 GB | Apache 2.0 | 2.6M | 1K |
| 7 | opt-125m | 125M · FP16 | ~1 GB | Custom (other) | 4.5M | 333 |
| 8 | Llama-3.2-1B-Instruct meta-llama | 1B · FP16 | ~3 GB | Llama 3.2 license | 1.8M | 1.9K |
| 9 | Qwen2.5-3B-Instruct Qwen | 3B · FP16 | ~8 GB | Custom (other) | 3.1M | 631 |
| 10 | Llama-3.2-3B-Instruct meta-llama | 3B · FP16 | ~8 GB | Llama 3.2 license | 995.2K | 2.9K |
| 11 | MiniCPM5-2B openbmb | 2B · FP16 | ~5 GB | Apache 2.0 | 1.1M | 1.8K |
| 12 | TinyLlama-1.1B-Chat-v1.0 TinyLlama | 1.1B · FP16 | ~3 GB | Apache 2.0 | 929.6K | 1.9K |
| 13 | SmolLM2-135M HuggingFaceTB | 135M · FP16 | ~1 GB | Apache 2.0 | 2.4M | 266 |
| 14 | DeepSeek-R1-Distill-Qwen-1.5B deepseek-ai | 1.5B · FP16 | ~4 GB | MIT | 784.6K | 1.7K |
| 15 | gemma-3-1b-it | 1B · FP16 | ~3 GB | Gemma license | 826.6K | 1.3K |
How we rank
Text-generation models with 4 billion parameters or fewer. We include models seen on Hugging Face in the last 14 days with at least 10,000 downloads in the last 30 days. Score = log10(30-day downloads) + ½ × log10(likes + 1), so popularity leads and likes break near-ties. We keep one entry per model name and leave out adult and uncensored builds. Memory is estimated from parameter count and the published file format (16-bit ≈ 2 bytes per parameter, 4-bit ≈ 0.5) plus 20% headroom; long contexts need more. A quantised copy of a bigger model may also fit — check the model page.