Ollama Now Runs Faster on Macs Thanks to Apple's MLX Framework

Tuesday March 31, 2026 3:22 am PDT by Tim Hardwick

Ollama, the popular app for running AI models locally on a computer, has released an update that takes advantage of Apple's own machine learning framework, MLX. The result is a hefty speed boost on Macs with Apple silicon.

According to Ollama, the new version processes prompts around 1.6 times faster (prefill speed) and nearly doubles the speed at which it generates responses (decode speed). Macs with M5-series chips are said to see the largest improvements, thanks to Apple's new GPU Neural Accelerators.

The update also includes smarter memory management, which should make AI-powered coding tools and chat assistants feel noticeably more responsive during extended use.

Ollama says the new performance boost should especially benefit macOS users who run personal assistants like OpenClaw or coding agents like Claude Code, OpenCode, or Codex.

The preview release is available to download as Ollama 0.19 – just make sure you have a Mac with more than 32GB of unified memory to run it. Support is currently limited to Alibaba's Qwen3.5, but Ollama says support for more AI models is planned.

Popular Stories

Apple Alerts Stores to iPhone Duo's Display Cracking Issue

Friday October 30, 2026 3:20 pm PDT by Joe Rossignol

In a memo shared with Apple Stores and Apple Authorized Service Providers this week, and later obtained by MacRumors, Apple provided technicians with guidance on how to diagnose and address the iPhone Duo's inner display cracking issue. If a customer reports the issue, Apple advised technicians to inspect the iPhone Duo and repair or replace it in accordance with their standard procedures....

• 229 comments

Apple Executive Comments on New iPad Mini Lacking 120Hz Display

Thursday October 29, 2026 1:47 pm PDT by Joe Rossignol

While the new iPad mini has an OLED display, the device does not support ProMotion, meaning the display's refresh rate is limited to 60Hz instead of 120Hz. ProMotion support would have allowed for the iPad mini's refresh rate to automatically adjust from 10Hz to 120Hz depending on the content being displayed, but evidently Apple has kept the feature exclusive to the iPad Pro among iPad...

• 243 comments

Apple Avoids Showing Pinched Fingers on iPad Mini Page in South Korea

Thursday October 29, 2026 11:52 am PDT by Joe Rossignol

In the U.S. and most other countries, Apple's website shows the new iPad mini being held by pinched fingers, but this is not the case in South Korea. The new iPad mini shown with pinched fingers on Apple's website in the U.S. On its South Korean website, Apple simply shows an image of the iPad mini on its own. The new iPad mini shown without pinched fingers on Apple's website in South Korea ...

• 94 comments

Top Rated Comments

neilpmas

8 months ago

This is going to be some serious cash flow incoming for Apple in this year.

I think this could be a major business for Apple - it’s way cheaper for a small business to buy a powerful Mac and run qwen 3.5 than pay for an enterprise license for a frontier model - and you don’t need to worry about privacy issues.

Score: 11 Votes (Like | Disagree)

RemedyRabbit

8 months ago

On device is definitely gonna be the future.

I can’t help but wonder if Apple looked ahead and foresaw this when developing the M series, or if they’ve lucked into it.

Score: 10 Votes (Like | Disagree)

Justin Cymbal

8 months ago

M-Series chips at work😎

Score: 7 Votes (Like | Disagree)

JMalone

8 months ago

As someone who downloads and experiments with everything possible…

There is a lot of delusion in this thread. Local language models below 100 billion parameters are quite useless. Even 100 billion parameters is considered the weak side. Fun to play with for a while but boredom and frustration sets in quickly.

So what happens is they want the next model…and then the next one…and then the next one…falsely believing their 16GB or 32GB machine will one day have the holy grail of small and powerful local language model.

But it doesn’t happen. The models keep growing and aside from being memory hungry the most important thing that makes them useable is memory bandwidth.

The top 5 language models in the world are all over a trillion parameters and what makes them useful and responsive is that they respond quickly and have GPU with over a terabyte of bandwidth.

Score: 6 Votes (Like | Disagree)