Local AI Scout · Local · News · Docs-only · Updated November 14, 2026
Short answer: The useful change this week is that Ripple Preview, the 65B coding model, is now in Ollama. Ollama v0.40.16, released November 9, 2026, adds it as ollama run ripple-preview, and also says Qwen3.8 Flash Next uses about 27 GiB less memory on Apple silicon ‡. We have not downloaded or run these builds. Comfortable hardware for Ripple is still in the region our existing guide gives: about 48GB or more, with 64GB the comfortable target.
Ripple Preview is now in Ollama
When we wrote last week's edition, Ripple-65B Preview was not in the Ollama library. Ollama v0.40.16, released November 9, 2026, added it. The release notes describe it as "a 65B dense model built for agentic coding," with this command ‡:
ollama run ripple-preview
The Ollama library page, which we opened on November 14, 2026, still lists the model from Ricursive Intelligence under the Apache 2.0 license, with download sizes of 39GB to 46GB. One thing to watch: the page blurb says "up to 1M tokens of context," while the model table on the same page lists a 2M context window ‡. We have not run it, so we can't tell you which you'll get.
A 39GB download is still a big-machine model. It does not make Ripple a 32GB-laptop option. For memory, the llama.cpp route and the full tag list, see our guide on how to run Ripple-65B locally.
Qwen3.8 Flash Next needs less memory on Apple silicon
The same v0.40.16 release notes say Qwen3.8 Flash Next now uses about 27 GiB less memory on Apple silicon ‡. That is Ollama's figure, not one we measured, and we are not turning it into a benchmark. If you have been holding off on this model because of memory, recheck what fits your Mac with how much memory a local AI model needs and our models table.
Only two items this week
We could not confirm a third local AI change on a primary source in time for this edition, so we are shipping two rather than padding it out.
From last week
For Kimi K3.1 Pro, MiMo V2.6 in Ollama 0.40.15 and the earlier llama.cpp-only Ripple note, see Local AI News, November 6, 2026. This edition replaces last week's note on Ripple and Ollama.
Frequently asked questions
Can I run Ripple in Ollama now?
Yes. Ollama v0.40.16, released November 9, 2026, added it as ollama run ripple-preview, and the Ollama library page still listed it when we checked on November 14, 2026. We have not run it.
How much memory do I need for Ripple?
Ollama lists download sizes of 39GB to 46GB. Our Ripple guide puts comfortable memory at about 48GB or more, with 64GB the comfortable target. Those are planning figures, not measurements.
Have you tested these models?
No. This edition is a documentation review only. We have not downloaded or run Ripple Preview or Qwen3.8 Flash Next.
Sources: Ollama v0.40.16 release notes, November 9, 2026: https://github.com/ollama/ollama/releases/tag/v0.40.16 · Ollama library, ripple-preview, opened November 14, 2026: https://ollama.com/library/ripple-preview
‡ The Ollama library page conflicts with itself on context length (1M in the blurb, 2M in the model table). The release-note wording and the Qwen3.8 memory figure come from Ollama's v0.40.16 notes, which we could not reopen on November 14, 2026; recheck them before relying on either.
Checked November 14, 2026 (Europe/London). Evidence: documentation review. We have not downloaded or run any model in this edition. Editorial policy.
Want the weekly roundup in your inbox? Sign up to the newsletter, or start with the Ripple guide.
