Model Size Visualizer
How big is 405 billion parameters compared to 7 billion? This tool makes the scale differences visible.
Logarithmic scale. Differences between small models are more visible.
Size comparison
| Model | Provider | Parameters | vs. smallest | Released |
|---|---|---|---|---|
| GPT-4 | OpenAI | 1,760B | 463.2x | 2023 |
| DeepSeek V3 | DeepSeek | 685B | 180.3x | 2024 |
| Llama 3.1 405B | Meta | 405B | 106.6x | 2024 |
| Mistral Large | Mistral | 123B | 32.4x | 2024 |
| Llama 3.1 70B | Meta | 70B | 18.4x | 2024 |
| Gemma 2 9B | 9.2B | 2.4x | 2024 | |
| Llama 3.1 8B | Meta | 8B | 2.1x | 2024 |
| Mistral 7B | Mistral | 7.3B | 1.9x | 2023 |
| Phi-3 Mini | Microsoft | 3.8B | 1x | 2024 |
What are parameters?
Parameters are the numerical values a model learns during training. Think of them as the model's "memory." More parameters generally means the model can store more knowledge and capture more subtle patterns in language.
But parameter count isn't everything. A 7B model trained on high-quality data with good techniques can outperform a 13B model trained poorly. Llama 3.1 8B, for example, beats many older 13B models on standard benchmarks.
The real question isn't "how many parameters?" but "how much capability per dollar?" That's where the LLM Cost Calculator comes in.