Burkov dismisses Jev's speed claims; suggests ONNX DeBERTa classifiers beat LLMs
MaziyarPanahi · x · 2026-09-21
Several people asked Andriy Burkov about a product called Jev; he called it "looks like BS": Jev claims to be much faster than asking an LLM to do the same task, but the advantage may only matter for niche cases like high-frequency trading — and it compares against regular LLMs, not diffusion LLMs. Maziyar Panahi added that encoders/GLi-models deserve mention: a DeBERTa classifier in ONNX format runs incredibly fast on CPU.
Related event: Burkov calls Jev's speed claims BS over weak LLM baseline(2 posts)→
More from Infra
- Inside SemiAnalysis: the research firm guiding a $1T AI infrastructure buildout — AccBalanced · 2026-09-21
- NEAR AI Cloud launches confidential inference on SayGm, both sides run in Intel TDX enclaves — bittingthembits · 2026-09-21
- Dual Intel Arc Pro B60s Eat Host RAM via GTT — a One-Line Level Zero Workaround — fobsthedev · 2026-09-21
- Running 2.8T-param Kimi K3 on a 16x GB10 Cluster: 30 tok/s Coding Decode — ciprianveg · 2026-09-21
- Prompt cache restore works, but reusing KDA/Mamba states hits illegal memory access — TheZachMueller · 2026-09-21
- 128k context with compaction beats raw 256k/1M for quantized models — Informal-Trouble2183 · 2026-09-21