Side by side
Compare two architectures
Pick any two models to see their diagrams, numbers and design choices next to each other.
Llama 3 (8B) versus Qwen3 (8B)
| Property | Llama 3 (8B) | Qwen3 (8B) |
|---|---|---|
| Total parameters | 8B | 8B |
| Active per token | 8B | 8B |
| Context window | 8K | 128K16× |
| KV cache / token | 128 KiB | 144 KiB1.1× |
| Decoder typeidentical | Dense | Dense |
| Attention | GQA with RoPE | GQA with QK-Norm |
| Layer mix | 32 GQA | 36 GQA |
| License | Llama 3 Community License Agreement | Apache License 2.0 |
| Released | April 18, 2024 | April 28, 2025 |
| Key detail | Pre-norm baseline; wider than OLMo 2 at a similar scale. | Reference Qwen3 dense stack with QK-Norm and 8 KV heads. |
Popular comparisons
- Llama 3 8B vs Qwen3 8B
- DeepSeek V3 vs Kimi K2
- GPT-OSS 120B vs GLM-4.5-Air
- DeepSeek V3.2 vs DeepSeek V4-Pro
- Qwen3 30B-A3B vs Qwen3 Next 80B-A3B
- Gemma 3 27B vs Gemma 4 31B
- GPT-2 XL 1.5B vs Llama 3 8B
- Mistral Large 3 vs DeepSeek V3
Agents: comparisons are client-side, so fetch two fact sheets directly, e.g. /api/models/<slug>.json.