Utopia Talk - Politics - Don’t be scammed by the Americans

Welcome to the Utopia Forums! Register a new account
The current time is Sun Nov 08 19:01:23 UTC 2026

Utopia Talk / Politics / Don’t be scammed by the Americans

Paramount
rank Sun Nov 08 14:57:41
Choose Chinese AI for less cost and better performance

Some Revolut engineers were spending up to $10,000 a month on AI tools. Now the company says it has cut its AI coding costs by 90%.

The fintech has moved much of its coding workload from Anthropic's Claude Code and OpenAI's Codex to GLM-5.3, an open AI model developed by Chinese company Z.ai.

Instead of paying for access through external providers, Revolut runs the model on 56 NVIDIA B200 GPUs, serving around 2,000 engineers.

According to Revolut's AI operations lead, the company now spends roughly $0.30 per million tokens, about one-tenth of what it previously spent on Anthropic and OpenAI.

The company also reports a 60% improvement in coding time and a 70% improvement in the time needed to deliver software changes.

Revolut still uses Claude Code, Codex and Gemini for certain tasks, but its own GLM system now handles much of the work.

https://ww...hare-7523411970876764160-WF6u/

Pillz
rank Sun Nov 08 15:35:51
Chinese advancements are purely in model architecture.

For data and training they are only now spinning up efforts. Everything else was distillation of US models.

Which means Chinese models are less 'complete', and always behind in capabilities.

Also, Chinese labs are starting to safetymaxxx models in areas like 'hacking', so it's possible in 12 months Chinese open weight models are as safety aligned as US ones, with 'bad' data never reaching the model to begin with.

Sad times.

Anyways, AI did end the world, just not in the skynet way.

Everyone is a slave now.

Average Ameriacn
rank Sun Nov 08 16:28:46
"runs the model on 56 NVIDIA B200 GPUs"
AI says that system costs several millions!?!?!

show deleted posts

Paramount rank	Sun Nov 08 14:57:41 Choose Chinese AI for less cost and better performance Some Revolut engineers were spending up to $10,000 a month on AI tools. Now the company says it has cut its AI coding costs by 90%. The fintech has moved much of its coding workload from Anthropic's Claude Code and OpenAI's Codex to GLM-5.3, an open AI model developed by Chinese company Z.ai. Instead of paying for access through external providers, Revolut runs the model on 56 NVIDIA B200 GPUs, serving around 2,000 engineers. According to Revolut's AI operations lead, the company now spends roughly $0.30 per million tokens, about one-tenth of what it previously spent on Anthropic and OpenAI. The company also reports a 60% improvement in coding time and a 70% improvement in the time needed to deliver software changes. Revolut still uses Claude Code, Codex and Gemini for certain tasks, but its own GLM system now handles much of the work. https://ww...hare-7523411970876764160-WF6u/
Pillz rank	Sun Nov 08 15:35:51 Chinese advancements are purely in model architecture. For data and training they are only now spinning up efforts. Everything else was distillation of US models. Which means Chinese models are less 'complete', and always behind in capabilities. Also, Chinese labs are starting to safetymaxxx models in areas like 'hacking', so it's possible in 12 months Chinese open weight models are as safety aligned as US ones, with 'bad' data never reaching the model to begin with. Sad times. Anyways, AI did end the world, just not in the skynet way. Everyone is a slave now.
Average Ameriacn rank	Sun Nov 08 16:28:46 "runs the model on 56 NVIDIA B200 GPUs" AI says that system costs several millions!?!?!
	show deleted posts

Your Name:
Your Password:
Your Message: