VQV Signal

AI AT WORK WATCH TECHNICAL

Revolut moves about 2,000 engineers to self-hosted GLM 5.3 as default coding model

According to a post on X by @tanayj, Revolut has switched its roughly 2,000 engineers to a self-hosted GLM 5.3 as their default coding model. The post says this cuts its coding AI spend by 10x compared with OpenAI and Anthropic. The same post points to a separate list of open models worth archiving this month, including DeepSeek v4.1 Flash, GLM 5.3 Flash and two Qwen 3.8 models, which drew over 200 likes. The provided snippet does not explain how the 10x figure was measured.

Source: Scobleizer News Pulse (X) · x.com Found via Robert Scoble’s Scobleizer News Pulse · a post on X, not independently verified Published 2026-11-15T02:40:00+00:00 Detected 2026-11-15T18:20:23+00:00
View original source

According to a post on X by @tanayj, Revolut has switched its roughly 2,000 engineers to a self-hosted GLM 5.3 as their default coding model. The post says this cuts its coding AI spend by 10x compared with OpenAI and Anthropic. The same post points to a separate list of open models worth archiving this month, including DeepSeek v4.1 Flash, GLM 5.3 Flash and two Qwen 3.8 models, which drew over 200 likes. The provided snippet does not explain how the 10x figure was measured.

AI-assisted summary based on the listed source.

Tanay Jaipuria reports that Revolut switched its roughly 2,000 engineers to a self-hosted GLM 5.3 as their default coding model, cutting its coding AI spend by 10x compared with OpenAI and Anthropic. Separately, Ahmad Osman's list of open models worth archiving this month (DeepSeek v4.1 Flash, GLM 5.3 Flash and...

If the claim holds, it shows a large company running an open model in-house for coding work at scale, as an alternative to paid services from OpenAI and Anthropic. Teams comparing self-hosting with paid APIs will want the method behind the 10x spend figure. The source is a single social post, and the material provided does not show that VQV or Revolut has confirmed the details.

Teams using AI at work may want to compare this against current productivity and review workflows.

Signal Strength 93% Technical label WATCH Public Interest 39 Category AI AT WORK 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 60 Practical Impact Score 8 Novelty Interest Score 70 Consequence Score 18 Curiosity Score 0 Shareability Score 55

VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).