Llama 3.3 70B Instruct
meta DENSE text Llama-3.3-Community 2024-12-06
Architecture
Total params
70 B
Active params
70 B
Layers
80
Context
128 k
Detailed specs
Hidden size
8192
FFN size
28672
Attention heads
64
KV heads
8
Head dim
128
Vocab size
128256
Attention type
gqa
Operator decomposition (per token)
Compatible hardware
— unknown AMD Instinct MI300A — unknown AMD Instinct MI300X — unknown AMD Instinct MI325X — unknown AMD Instinct MI355X — unknown Apple M4 Max Neural Engine — unknown AWS Inferentia 2 — unknown AWS Trainium 2 — unknown 壁仞 BR100 — unknown 壁仞 BR104 — unknown 寒武纪 MLU370-X8 — unknown 寒武纪 思元 590 — unknown Cerebras WSE-3 — unknown 燧原 云燧 T21 — unknown Etched Sohu — unknown Google TPU v5p — unknown Google TPU Trillium (v6e) — unknown Groq LPU (TSP v1) — unknown 昇腾 910B — unknown 昇腾 910C — unknown 昇腾 950 — unknown 海光 DCU K100 — unknown 海光 DCU Z100 — unknown 天数智芯 天垓 100 — unknown Intel Gaudi 2 — unknown Intel Gaudi 3 — unknown 沐曦 曦云 C500 — unknown 摩尔线程 MTT S4000 🟢 measured NVIDIA A100 SXM4 80GB — unknown NVIDIA B200 SXM 180GB — unknown NVIDIA B300 SXM 288GB — unknown NVIDIA GB200 NVL72 — unknown NVIDIA GB300 NVL72 — unknown NVIDIA H100 SXM5 80GB — unknown NVIDIA H200 SXM 141GB — unknown NVIDIA L40S — unknown NVIDIA R200 SXM (Vera Rubin) — unknown 平头哥 含光 800 — unknown SambaNova SN40L — unknown Tenstorrent Wormhole n300