moonshotai/Kimi-K2.5
Image-Text-to-Text ā¢ 1.1T ā¢ Updated ā¢ 5.56M ā¢ ā¢ 2.46k
Jonna Matthiesen
JonnaMat
AI & ML interests
None yet
Recent Activity
posted an update about 7 hours 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.
updated a model about 8 hours ago
embedl/Qwen3-1.7B-FlashHead-W4A16 updated a model about 8 hours ago
embedl/gemma-3-1b-it-FlashHead-W4A16Organizations
liked a
model 30 days ago
liked a
model about 1 month ago
liked a
Space about 1 month 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¢ 87 ā¢ 4
embedl/gemma-3-1b-it-FlashHead-W4A16
1B ā¢ Updated ā¢ 138 ā¢ 3
embedl/Llama-3.2-3B-Instruct-FlashHead
3B ā¢ Updated ā¢ 69 ā¢ 4
embedl/gemma-3-1b-it-FlashHead
1.0B ā¢ Updated ā¢ 130 ā¢ 3
embedl/Llama-3.2-1B-Instruct-FlashHead-W4A16
2B ā¢ Updated ā¢ 71 ā¢ 6
embedl/Qwen3-1.7B-FlashHead-W4A16
2B ā¢ Updated ā¢ 105 ā¢ 3
embedl/Qwen3-1.7B-FlashHead
2B ā¢ Updated ā¢ 70 ā¢ 3
embedl/Llama-3.2-1B-Instruct-FlashHead
1B ā¢ Updated ā¢ 51 ā¢ 4