Ripple-65B open-weight model runs at 7 tok/s locally on a 64GB Mac mini
DZhang50 · x · 2026-11-06
Ripple-65B Preview from RicursiveAI is a 65B dense open-weight model (Apache 2.0, pretrained from scratch). A user tested the community MLX 4-bit build locally on a 64GB M5 Pro Mac mini:
- Steady 7 tok/s from 512 to 32K tokens
- Peaks at 40.5 GB memory, leaving headroom on 64GB
- Passed 53/53 local tests, hidden facts recalled out to 64K context
- Runs about as fast as the chip allows for a dense 65B
The community shipped an MLX build within a day of release — "a good day for owning your intelligence."
More from Infra
- The biggest AI bottleneck is now communication about data centers, argues Matt Slotnick — matt_slotnick · 2026-11-06
- SGLang Summit 2026 unveils RL and post-training systems agenda for Nov 12-13 — BanghuaZ · 2026-11-06
- vLLM open-sources System Duo: a fast/slow semantic router that only calls deep reasoning when needed — vllm_project · 2026-11-06
- 72 tok/s code generation on a 12GB RTX 3060: Reddit user's custom Beellama fork beats Ternary Bonsai — SeanHighness · 2026-11-06
- Blogger Proposes Data Centers Win Locals Over With Free ChatGPT Pro and Waste Heat — flowersslop · 2026-11-06
- Voters reject data centers in 40 of 41 ballot measures; Musk eyes orbital data centers — PeterDiamandis · 2026-11-06