Free Speed Boost: Ollama Makes Local AI Up to 90% Faster on Macs
Ollama — the easiest way to run AI models on your own computer — just made Google's Gemma 4 model dramatically faster on Apple machines. No settings, no trade-offs: update and it's quicker. Here's how the clever trick behind it works.
Every so often, software gets better in the nicest possible way: you update it, and it’s simply faster — nothing to configure, nothing to pay, no catch. That’s what happened this week for people who run AI models on their own Macs. Ollama, a free tool that makes running AI models locally about as easy as installing an app, shipped version 0.31 — and with it, Google’s Gemma 4 model now generates text up to 90% faster on Apple computers.
Quick orientation if this world is new to you: instead of chatting with an AI in the cloud (like ChatGPT), you can download a free model and run it directly on your computer. Your questions never leave your machine, it works offline, and it costs nothing per use. The traditional catch has been speed — local models often feel sluggish next to their cloud cousins. Which is why updates like this one matter.
The trick behind the speedup is called multi-token prediction, and it’s genuinely clever. Normally a model produces text one token at a time (a token is a chunk of a word — “under-stand-ing” might be three). Now, a small, fast “draft” model sprints ahead and proposes the next several tokens, and the big model just checks the batch in one go — keeping what’s right, redoing what’s wrong. Like a junior writer drafting sentences a senior editor approves in bulk. The output is identical to what the big model would have written alone; it just arrives sooner. Predictable text like computer code benefits most (that’s where the 90% figure comes from); free-flowing conversation lands closer to 20–40% faster, which is still very noticeable.
Under the hood there are more speed gains on Apple’s newest chips, and Ollama says Gemma 4 is just the first model to get this treatment — more are coming.
What this means for you: If you already use Ollama with Gemma 4, run the update and enjoy — you’re done. If you’ve been curious about trying AI on your own computer but assumed it would be painfully slow: that assumption ages badly, fast. A recent Mac and a free evening are enough to try it — and updates like this keep quietly closing the gap with the cloud.
Sources
Scammers Found a New Trick: Registering the Web Addresses AI Makes Up
AI chatbots sometimes invent web addresses that don't exist — and researchers discovered they invent the same ones over and over. Now attackers are buying those addresses and building fake sites there, waiting for AI users to arrive. Here's how to not be one of them.
