Runtimes & apps Published:
Ollama v0.40.16 cuts Qwen3.8 Flash Next's memory use on Apple silicon by about 27 GiB

What happened
Ollama v0.40.16, released on November 9, says Qwen3.8 Flash Next now uses about 27 GiB less memory on Apple silicon. The release adds Ricursive's Ripple Preview coding model, makes `ollama launch claude` always route Claude Code to the model you selected, and explains in `ollama launch` when a cloud model has been retired. Its OpenAI-compatible `/v1/responses` endpoint now respects the `truncation` setting and no longer truncates silently by default, `/v1/models` returns an empty list instead of `null` when nothing is installed, and Windows on ARM gains CUDA 13.5 support.
Why it matters locally: Qwen3.8 Flash Next is one of the largest models people run on a single Mac, so freeing about 27 GiB leaves more room for long context or a second model on the same machine.
Three ways to read it
For the market
Software gains like this change what a given Mac can do without new hardware, which makes high-memory Apple machines look like a better buy for running large MoE models. It also puts pressure on other local runtimes to match Ollama's Apple silicon memory handling.
In plain words
Ollama is a free app that runs AI models on your own computer. This update makes one of the biggest models it supports take noticeably less memory on a Mac, like repacking a suitcase so the same clothes leave space for more.
For practitioners
Qwen3.8 Flash Next stores 180B parameters (a 125B MoE with 6B active plus n-gram embeddings), so it still needs a high-memory Mac; after upgrading, recheck how much context fits and whether you can now keep an embedding or coding model loaded alongside it. If your apps call `/v1/responses`, set `truncation` explicitly, since the default no longer cuts long inputs silently, and update any client code that treated a `null` from `/v1/models` as a special case.
Sprout's insight
Sprout compares this with other brands and with where real applications are heading. Free for members.
Sign in free to read it →Subscribe to Sprout's weekly
One email a week: the top stories and a taste of Sprout’s insights. Unsubscribe any time.
Source
Read the official source: Ollama v0.40.16 release notes (GitHub) ↗Summaries are written by OnPrem World from the official source. We link to the original and never republish it.
On this site
More news
- NVIDIA and Microsoft bring Nemotron 3.5 and RTX PRO 6000 Blackwell support to Azure Local
- Microsoft publishes Aion 1.0 Instruct, the 2.3B model that becomes Windows' built-in AI by end of January 2027
- RTX Spark compact desktops from ASUS, Dell and MSI go on sale with up to 128 GB of unified memory
- NVIDIA reports $115.8 billion quarter; highlights include a deal to acquire Hugging Face and a 64GB DGX Spark
- Open WebUI v0.13.4 moves personal connections into Tether, sandboxes agents on your own computer and runs local MCP servers