For the very hardest reasoning tasks GPT-5 and Opus are still ahead, no argument there. But what we see in practice is customers dropping in an open-source model and getting very similar results on 80-90% of real-world use cases — with significant cost savings and end-to-end UK data residency (which matters a lot for our enterprise and institutional customers).
And on the consumer-hardware point: these are Blackwell GPUs in a UK datacentre, in a token factory architecture.
This is very disingenuous, I've been deploying local models to enterprise across a variety of use cases and the optimisation overhead and prompt engineering required to get good performance is huge. Let alone comparative perf to frontier models.