Ornith-1.5-122B Arrives With a Day-Zero Quant for One Spark
Open weights under the MIT license; @MiaAI_lab posted an EXL3-EAQ v1 quant for one DGX Spark or RTX PRO 6000.
A new 122B open-weight model, quantized the same morning.
Ornith released Ornith-1.5-122B, which it bills as “No Tradeoff Between Strong and Fast.” In its announcement, quoted by @MiaAI_lab, Ornith says the model offers open weights under the MIT license and runs on a single NVIDIA DGX card.
@MiaAI_lab congratulated the team and posted a day-zero EXL3-EAQ v1 quant of the model for one DGX Spark or RTX PRO 6000, available for download.
“You’ll be surprised.”
About two hours later she followed up: “I’m actually blown away by how well Ornith-1.5-122B performs.” She says her 100-HTML experiment with the model is coming soon.
Her 100-HTML test asks a model for 100 HTML files. In this morning’s edition she reported MiniMax M3.1 Flash’s attempt, with some broken files.
Sources
- @MiaAI_lab on X · Oct 20, 8:03 AM PT
- @MiaAI_lab on X · Oct 20, 8:05 AM PT
- @MiaAI_lab on X · Oct 20, 10:05 AM PT