Summary
According to a post on X by @RedHat_AI, Red Hat ran its own engineering organization on open-weight models as "customer zero," using a Praxis gateway, llm-d and vLLM. The post says the setup processed more than 1 trillion tokens in 30 days. It also says usage shifted from a 70/30 split favoring commercial frontier APIs to a 30/70 split favoring internally hosted open weights. A reshared post from @Saboo_Shubham_ highlighted an open-source project for running agents, but the snippet cuts off before details.
AI-assisted summary based on the listed source.
What happened
A reshared @RedHat_AI post said Red Hat ran its engineering org on open-weight models as "customer zero" using a Praxis gateway, llm-d and vLLM, processing more than 1 trillion tokens in 30 days, with usage flipping from 70/30 in favor of commercial frontier APIs to 30/70 toward internally hosted open weights. A...
Why it matters
Teams weighing self-hosted open-weight models against commercial frontier APIs get a concrete internal example, along with the tools named in the post. Platform and infrastructure engineers using vLLM or llm-d may want to watch for fuller technical details from Red Hat. These figures come from a social post and have not been independently checked by VQV.
Signal Intelligence
Signal Strength 83%
Technical label WATCH
Public Interest 48
Category OPEN SOURCE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 51
Practical Impact Score 28
Novelty Interest Score 70
Consequence Score 42
Curiosity Score 32
Shareability Score 57
Why this is here
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).