A repost of @baseten (107 reposts) said it built an inference engine for DeepSeek V4.1-Flash from scratch in Rust and CUDA, reaching about 1,740 tokens per second decode on 4xB300 and about 2,060 on 8xB300. A repost of @hxiao described gemini-4-carbon hill...
VQV Signal
Fast inference builds: a from-scratch DeepSeek engine and a 15-hour kernel loop
A repost of @baseten (107 reposts) said it built an inference engine for DeepSeek V4.1-Flash from scratch in Rust and CUDA, reaching about 1,740 tokens per second decode on 4xB300 and about 2,060 on 8xB300. A repost of @hxiao described gemini-4-carbon hill...
A repost of @baseten (107 reposts) said it built an inference engine for DeepSeek V4.1-Flash from scratch in Rust and CUDA, reaching about 1,740 tokens per second decode on 4xB300 and about 2,060 on 8xB300. A repost of @hxiao described gemini-4-carbon hill-climbing a kernel for 15 hours in a simple...
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
VQV surfaced this signal because it is recent, relevant to LLM Inference, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.