Tencent fits a 35-billion-parameter model on an ordinary gaming PC.
Tencent says it cut the model's memory use from about 30 gigabytes to under 10. On a Windows PC with 8 gigabytes of graphics memory it makes about 50 word pieces a second. The trick shipped Wednesday in its Marvis desktop assistant, and 16-gigabyte machines follow in November.
Why it mattersA model this size on a home PC means your data never leaves the machine. It also means no monthly fee and no cut-off when the internet drops.
▶ Related video: LLM Quantization Explained
Read in the app — free, in 9 languages
Related stories
Amazon lets sellers hand their shop to Claude in 60 seconds.2026-09-25AWS says its free agent wrapper runs Claude at a quarter the price.
2026-09-24An AI agent made Anthropic's own apps three times faster.
2026-09-24Meta's assistant packed up its own machine and mailed it out.
2026-09-23Claude Code skips your instructions file when telemetry is off.
2026-09-23