Comparisons of Small Open Source AI Models (4B-40B)
Open source AI models with less than 40 billion parameters. Models are considered Open Source (also commonly referred to as open weights) where their weights are accessible to download. This allows self-hosting on your own infrastructure and enables customizing the model such as through fine-tuning. Click on any model to see detailed metrics. For more details including relating to our methodology, see our FAQs.
Qwen3 32B (Reasoning) and QwQ-32B are the highest intelligence Small open source models, defined as those with 4B-40B parameters, followed by Qwen3 14B (Reasoning) & Qwen3 30B A3B (Reasoning).
Highlights
Intelligence
Artificial Analysis Intelligence Index; Higher is better
Total Parameters
Trainable parameters in billions
Further details
| Weights | Provider Benchmarks | |||||||
|---|---|---|---|---|---|---|---|---|
Qwen3 32B (Reasoning) Alibaba | 59 | 32.8B | 128k | $2.6 | 61 | 🤗 |   +6 more | View |
QwQ 32B Alibaba | 58 | 32.8B | 131k | $0.5 | 136 | 🤗 |   +5 more | View |
Qwen3 14B (Reasoning) Alibaba | 56 | 14.8B | 128k | $1.3 | 66 | 🤗 | | View |
Qwen3 30B A3B (Reasoning) Alibaba | 56 | 30.5B (3.3B active at inference time) | 128k | $0.8 | 89 | 🤗 |   +4 more | View |
 Magistral Small Mistral | 55 | 23.6B | 128k | $0.8 | 195 | 🤗 |  | View |
 DeepSeek R1 0528 Qwen3 8B DeepSeek | 52 | 8.19B | 128k | $0.1 | 91 | 🤗 |  | View |
DeepSeek R1 Distill Qwen 32B DeepSeek | 52 | 32B | 128k | $0.3 | 28 | 🤗 |  | View |
Qwen3 8B (Reasoning) Alibaba | 51 | 8.19B | 128k | $0.7 | 99 | 🤗 | | View |
DeepSeek R1 Distill Qwen 14B DeepSeek | 49 | 14B | 128k | $0.2 | 83 | 🤗 | | View |
Qwen3 4B (Reasoning) Alibaba | 47 | 4.02B | 32.0k | $0.4 | 105 | 🤗 | | View |
 Reka Flash 3 Reka AI | 47 | 21B | 128k | $0.3 | - | 🤗 | | View |
 Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning) NVIDIA | 45 | 4.51B | 128k | - | - | 🤗 | - | View |
Qwen3 32B Alibaba | 44 | 32.8B | 128k | $1.2 | 61 | 🤗 | | View |
QwQ 32B-Preview Alibaba | 43 | 32.8B | 32.8k | $0.7 | 60 | 🤗 | | View |
Qwen3 30B A3B Alibaba | 43 | 30.5B (3.3B active at inference time) | 128k | $0.3 | 89 | 🤗 | | View |
Mistral Small 3.2 Mistral | 42 | 24B | 128k | $0.1 | 123 | 🤗 | | View |
Qwen3 14B Alibaba | 41 | 14.8B | 128k | $0.6 | 66 | 🤗 | | View |
 Phi-4 Microsoft Azure | 40 | 14B | 16.0k | $0.2 | 22 | 🤗 | | View |
 Gemma 3 27B Instruct Google | 38 | 27.4B | 128k | - | 45 | 🤗 | | View |
Qwen2.5 Instruct 32B Alibaba | 37 | 32B | 128k | $0.1 | - | 🤗 | | View |
Qwen3 8B Alibaba | 37 | 8.19B | 128k | $0.3 | 101 | 🤗 | | View |
Qwen2.5 Coder Instruct 32B Alibaba | 36 | 32B | 131k | $0.1 | 48 | 🤗 |  +1 more | View |
Mistral Small 3.1 Mistral | 35 | 24B | 128k | $0.1 | 182 | 🤗 | | View |
Mistral Small 3 Mistral | 35 | 24B | 32.0k | $0.1 | 162 | 🤗 | | View |
Qwen3 4B Alibaba | 35 | 4.02B | 32.0k | $0.2 | 106 | 🤗 | | View |
Devstral Mistral | 34 | 23.6B | 256k | $0.1 | 130 | 🤗 | | View |
DeepSeek R1 Distill Llama 8B DeepSeek | 34 | 8B | 128k | $0.0 | 56 | 🤗 | | View |
Gemma 3 12B Instruct Google | 34 | 12.2B | 128k | $0.1 | - | 🤗 | | View |
 Solar Mini Upstage | 33 | 10.7B | 4.10k | $0.1 | 93 | 🤗 | | View |
Gemma 2 27B Google | 32 | 27.2B | 8.19k | $0.8 | - | 🤗 | | View |
 DeepHermes 3 - Mistral 24B Preview Nous Research | 30 | 24B | 32.0k | - | - | 🤗 | - | View |
Gemma 3n E4B Instruct Google | 28 | 8.39B (4.0B active at inference time) | 32.0k | $0.0 | 70 | 🤗 | | View |
Mistral Small (Sep '24) Mistral | 27 | 22B | 32.8k | $0.3 | 121 | 🤗 | | View |
Gemma 3n E4B Instruct Preview (May '25) Google | 27 | 8.39B (4B active at inference time) | 32.0k | - | - | 🤗 | - | View |
Phi-4 Multimodal Instruct Microsoft Azure | 27 | 5.6B | 128k | - | 22 | 🤗 | | View |
Qwen2.5 Coder Instruct 7B Alibaba | 27 | 7.62B | 131k | - | - | 🤗 | - | View |
Gemma 3 4B Instruct Google | 25 | 4.3B | 128k | $0.0 | - | 🤗 | | View |
 Llama 3.2 Instruct 11B (Vision) Meta | 25 | 11B | 128k | $0.1 | 62 | 🤗 |  | View |
Phi-3 Medium Instruct 14B Microsoft Azure | 25 | 14B | 128k | $0.3 | 53 | 🤗 | | View |
Llama 3.1 Instruct 8B Meta | 24 | 8B | 128k | $0.1 | 225 | 🤗 |  +16 more | View |
Pixtral 12B (2409) Mistral | 23 | 12B | 128k | $0.1 | 41 | 🤗 | | View |
Ministral 8B Mistral | 22 | 8B | 128k | $0.1 | 201 | 🤗 | | View |
Gemma 2 9B Google | 22 | 9B | 8.19k | $0.1 | - | 🤗 | +1 more | View |
Llama 3 Instruct 8B Meta | 21 | 8B | 8.19k | $0.1 | 82 | 🤗 |  +4 more | View |
DeepSeek Coder V2 Lite Instruct DeepSeek | 20 | 16B (2.4B active at inference time) | 128k | - | - | 🤗 | - | View |
Codestral (May '24) Mistral | 20 | 22B | 32.8k | $0.3 | - | 🤗 | | View |
 Aya Expanse 32B Cohere | 20 | 32B | 128k | $0.8 | 124 | 🤗 | | View |
DeepSeek LLM 67B Chat (V1) DeepSeek | 20 | 7B | 4.10k | - | - | 🤗 | - | View |
Llama 2 Chat 13B Meta | 20 | 13B | 4.10k | - | - | 🤗 | - | View |
 OpenChat 3.5 (1210) OpenChat | 20 | 7B | 8.19k | $0.1 | 50 | 🤗 | | View |
Mistral NeMo Mistral | 20 | 12B | 128k | $0.1 | 158 | 🤗 | +1 more | View |
DeepHermes 3 - Llama-3.1 8B Preview Nous Research | 16 | 8B | 128k | - | - | 🤗 | - | View |
Aya Expanse 8B Cohere | 16 | 8B | 8.00k | $0.8 | 176 | 🤗 | | View |
Command-R (Aug '24) Cohere | 15 | 32B | 128k | $0.3 | 75 | 🤗 | | View |
Command-R (Mar '24) Cohere | 15 | 35B | 128k | $0.8 | 169 | 🤗 | | View |
Codestral-Mamba Mistral | 14 | 7B | 256k | $0.3 | - | 🤗 | | View |
Mistral 7B Instruct Mistral | 10 | 7B | 8.19k | $0.3 | 124 | 🤗 | +3 more | View |
Llama 2 Chat 7B Meta | 8 | 7B | 4.10k | $0.1 | 132 | 🤗 | | View |