inference
The Canadian company designs chips built to run just one model.
"You can't be subject to the refusal of one model. There's a lot more design space here…”
DeepMind turned to a 2022 paper to speed up its open source Gemma 4 models.
DeepSeek V4 and Qwen3.6-27B, both out last week, put pressure on Western models – especially for pricing.
Goldman Sachs: "Industry is countering by redistributing value [via] open source distilled models and the focus on roadmaps weighted to proprietary SLMs..."