Llama
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Llama-2 70B
97ASR
11
Llama 2 13B
96Attack Success Rate (ASR)
11
Llama 1.7B 3.2
28.3Energy Reduction (Iso-Time)
11
LLaMA 8B 3.1
1.99Mean Perplexity
10
LLaMA 2 7B
2.75Mean Perplexity (PPL)
10
LLaMA 7B
2.78Mean PPL
10
Llama 3.2 Google Pixel 9 Pro XL 1B (test)
530.02Prefill Throughput (min) (tokens/sec)
10
Llama 8B 3.1
1AUC
10
Llama 7b 2
0Avg KLD
10
LLaMA 8B 8K context length 3.1
159Theoretical Compute (TFLOPs)
10
LLaMA-350M pre-training (val)
2.707Validation Loss
10
LLaMA-30B
18.59Memory Footprint (GB)
10
LLaMA-13B
12.34Memory Cost (GB)
10
LLaMA 7B
8.76Memory Cost (GB)
10
Llama-2-70B
27.2Latency (ms)
10
Llama3 GCG Attack (test)
9.61ASR
10
Llama base variant (S=1024) on B200 3.1-8B
178.1Latency (ms)
9
Llama-3.1-8B base (train)
99.1Latency (ms)
9
LLaMA-2
98.5BERT Score
9
LLaMA
29.5Tuning Time (s)
9
Llama-2-7b chat-hf (test)
0Bit Error Rate (BER)
9
Llama-3.1-8B-Instruct 128K input length
14.5TTFT (s)
9
Llama 8B Instruct 112K input length 3.1
12.7Time-to-First-Token (s)
9
Llama-3.1-8B-Instruct 96K input length
10.9TTFT (s)
9
Llama-3.1-8B-Instruct 80K input length
9.2TTFT (s)
9