embedl/Cosmos-Reason2-2B-W4A16-Edge2-FlashHead
Image-Text-to-Text ā¢ 2B ā¢ Updated ā¢ 1.23k ā¢ 7
Jonna Matthiesen
JonnaMat
AI & ML interests
None yet
Recent Activity
updated a Space 2 days ago
embedl/Edge-Inference-Benchmarks liked a model 7 days ago
embedl/Cosmos-Reason2-2B-W4A16-Edge2-FlashHead posted an update 8 days ago
ā” FlashHead benchmarks for Llama 3.2, Gemma 3, and Qwen3 are now on https://huggingface.co/spaces/embedl/Edge-Inference-Benchmarks !
These are some of the models used in the FlashHead paper - now easier to explore and compare interactively.
š Jetson AGX Thor (tok/s, batch=1):
- Llama-3.2-1B: 77 ā 285 (FlashHead+W4A16, 3.7x)
- Llama-3.2-3B: 34 ā 112 (3.3x)
- Gemma-3-1B: 79 ā 153 (1.9x)
- Qwen3-1.7B: 49 ā 189 (3.8x)
- Qwen3-0.6B: 140 ā 177 (1.3x)
ā
Accuracy matches baseline on MMLU-Pro, IFEval, BBH, TruthfulQA, GSM8K.
Organizations
liked a
model 7 days ago
liked a
model about 1 month ago
liked a
Space about 2 months ago
liked a
model about 2 months ago
liked 8
models 4 months ago
embedl/Llama-3.2-3B-Instruct-FlashHead-W4A16
4B ā¢ Updated ā¢ 101 ā¢ 4
embedl/gemma-3-1b-it-FlashHead-W4A16
1B ā¢ Updated ā¢ 148 ā¢ 3
embedl/Llama-3.2-3B-Instruct-FlashHead
3B ā¢ Updated ā¢ 84 ā¢ 4
embedl/gemma-3-1b-it-FlashHead
1.0B ā¢ Updated ā¢ 138 ā¢ 3
embedl/Llama-3.2-1B-Instruct-FlashHead-W4A16
2B ā¢ Updated ā¢ 81 ā¢ 6
embedl/Qwen3-1.7B-FlashHead-W4A16
2B ā¢ Updated ā¢ 140 ā¢ 3
embedl/Qwen3-1.7B-FlashHead
2B ā¢ Updated ā¢ 113 ā¢ 3
embedl/Llama-3.2-1B-Instruct-FlashHead
1B ā¢ Updated ā¢ 72 ā¢ 4