phi-3-small 7.4B
phi-3-small 7.4B is an open-weights language model from Microsoft, released April 23, 2024. It has 7.4B parameters (dense). On the Epoch Capabilities Index it scores 116.5, rank 192 of 256 rated models.
Check it against your own PC →
Can you run phi-3-small 7.4B locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q6_K, split GPU / CPU, roughly 7 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs, slowly | Q6_K (≈ lossless) | 5.2 GB | 3.6 GB | ~7 |
| 8 GB VRAM | Runs on GPU | IQ4_XS (good) | 6.7 GB | – | ~40 |
| 12 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 8.8 GB | – | ~34 |
| 16 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 8.8 GB | – | ~47 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 8.8 GB | – | ~85 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 8.8 GB | – | ~150 |
Benchmark results
| Benchmark (Epoch AI) | Score |
|---|---|
| MMLU | 67.6% |
Links
Models with a similar score
Gemini 1.5 Flash (May 2024) · GPT-4 (Jun 2023) · Mistral Large · Claude 3 Sonnet · Qwen2.5-Coder-32B · Mixtral 8x22B · Gemma 2 27B · Claude 3 Haiku
More from Microsoft
MAI-Cyber-1.1-Flash · MAI-Base-2 · MAI-Cyber-1-Flash · MAI-Code-1.1-Flash · Mage VL · MAI-Thinking-1 · MAI-Code-1-Flash · Phi-4