Model overview · Llama
Llama 3 8B
Generated from Architecture IR 0.1.0 · revision 7784fbf634
· report a correction
Overview
Fact sheet
- Layers
- 32
- Hidden size
- 4096
- Attention heads
- 32 · KV 8
- Vocab
- 128,256
- Context
- 8K
- Params
- 8B total
Architecture Genome
One column per layer — top: attention, bottom: FFN/MoE.
Layer
attention (GQA/MLA) linear / recurrent attention MoE dense FFN
Evidence
Every visible number resolves to one of these claims. Inferred or unknown values are never shown as fact.
| Claim | Value | Status | Source |
|---|---|---|---|
facts.num_hidden_layers | 32 | • reported | hf_config · $.num_hidden_layers |
facts.hidden_size | 4096 | • reported | hf_config · $.hidden_size |
facts.num_attention_heads | 32 | • reported | hf_config · $.num_attention_heads |
facts.vocab_size | 128256 | • reported | hf_config · $.vocab_size |
facts.context_tokens | 8192 | • reported | hf_config · $.max_position_embeddings |
facts.total_params | 8030000000 | • reported | model_card · parameter count |
facts.num_key_value_heads | 8 | • reported | hf_config · $.num_key_value_heads |
topology.attention_groups[0] | "32 GQA layers" | • reported | hf_config · layers config / architecture table |
topology.position_encoding.type | "rope" | • reported | hf_config · rope settings |
facts.active_params | 8030000000 | = derived | derived · dense or activated experts + attention params |
Compare
Compare this model with any other catalog model (see catalog).