BREAKING: Revolut moved ~2,000 engineers onto self-hosted GLM for coding and pushed Claude Code and Codex to the elimination list. Per David Abitbol, who leads AI ops there: 56 Nvidia B200s, a blended 30 cents per myriad tokens, approximately a tenth of what they paid Anthropic and OpenAI. Engineering guide period improved 70%. Here's the part that matters: the gains showed up whenever tokens stopped being rationed. Engineers discontinue preservation AI for the difficult tickets and pointed it at everything. Frontier models rotate into the exception, open weights rotate into the default. That's a pricing issue for all lab heading to IPO. h/t @OliverMolander Would you move your squad off frontier APIs if the invoice dropped 90%?
