Llama
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Llama 8B Instruct 64K input length 3.1
7.2TTFT (s)
9
Llama 8B Instruct 48K input length 3.1
5.2TTFT (s)
9
Llama 8B Instruct 32K input length 3.1
3.4TTFT (s)
9
LLaMA Vision 3.2
89.8ASR
9
Llama 4
0.965ASR
9
Llama 3.1
0.97ASR
9
Llama Instruct MLP features 3.1
45.8Score
9
Llama Instruct MLP features 3.1
89.8Score
9
Llama (S=4096) on B200 3.1-8B (base variant)
202Per-Step Latency (ms)
8
Llama S=2048 on B200 3.1-8B (base variant)
178.7Per-step Latency (ms)
8
Llama-3-8B 500 Mbps
1.485Prefill Latency (s)
8
Llama-3-8B 200 Mbps
1.541Prefill Latency (s)
8
Llama-3-8B 100 Mbps
1.545Prefill Latency (s)
8
Llama-3-8B 50 Mbps
1.547Prefill Latency (s)
8
Llama-3-8B 20 Mbps
1.549Prefill Latency (s)
8
Llama-3-8B 10 Mbps
1.563Prefill Latency (s)
8
LLaMA Original Languages 3.1-8B
80.37Generalization Score
8
LLaMA Evaluation Expanded Languages 3.1-8B
69.37Overall Score
8
Llama 3.1
71Faithfulness
8
Llama2 Average 7B
96.9F1 Score
8
Llama 3.2 3B
31Upload Size (MB)
8
Llama Instruct 3.1-8B (internal harness)
6,991Throughput (TPS)
8
Llama 70B Emulation 3.3
9.3Time Reduction
8
Llama MLP SAE 3.1
55.5Faithfulness Score
8
Llama Residual SAE features 3.1
37.2Score
8