Llama 3-8B
Llama 3-8B is an open-weights language model from Meta, released April 18, 2024. It has 8B parameters (dense). On the Epoch Capabilities Index it scores 109.4, rank 205 of 256 rated models.
Check it against your own PC →
Can you run Llama 3-8B locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q6_K, split GPU / CPU, roughly 6 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 | 4.2 GB | ~6 |
| 8 GB VRAM | Runs on GPU | IQ4_XS (good) | 7.0 GB | – | ~37 |
| 12 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 9.4 GB | – | ~32 |
| 16 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 9.4 GB | – | ~44 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 9.4 GB | – | ~79 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 9.4 GB | – | ~140 |
Strengths by category
ModelTape's 0–100 fit on Epoch AI benchmark results.
| Category | Score | Based on |
|---|---|---|
| Math | 7.9 | MATH level 5, OTIS Mock AIME 2024-2025 |
| Knowledge | 23.3 | GPQA diamond, MMLU |
Benchmark results
| Benchmark (Epoch AI) | Score |
|---|---|
| MMLU | 58.4% |
| MATH level 5 | 6.1% |
| OTIS Mock AIME | 1.8% |
| GPQA Diamond | 1.5% |
Links
Models with a similar score
Mixtral 8x7B · GPT-3.5 Turbo (Jan 2024) · GPT-3.5 Turbo (Nov 2023) · Gemma 2 9B · Llama 3.1-8B · phi-3-mini 3.8B · Stable Beluga 2 · Qwen2.5-Coder (7B)
More from Meta
Muse Ember · Muse Spark Open · Muse Wonder · Muse Wonder Full · Muse Spark 1.3 · Muse Spark 1.3 Contributor · Muse Glimmer · Muse Glimmer 30B assistant