Uploaded November 2026 | Updated November 2026, 3 hours ago
🖥️ Get the Olares One $600 off with our link (Black Friday Deal) -tinyurl.com/olares600
Save $300 after black Friday
💙⚡Become a member – tinyurl.com/blmember
If you're thinking about buying a Mac Studio for AI, watch this first. Olares sponsored this video and sent us an Olares One, a small personal AI computer with NVIDIA's laptop RTX 5090 and 96 GB of memory, so we raced it against the $11,299 Mac Studio M5 Ultra we just bought. Through our link the Olares One is $600 off, which brings it to $4,399. By the end you'll know which one is faster for local image and video models, what each one costs to buy and to upgrade, and which kind of work suits each.
We open it up to show the RTX 5090 laptop chip, the 24-core Intel Core Ultra 9, the two memory sticks and the free SSD slot, and explain why its 24 GB of graphics memory and 96 GB of system memory behave so differently from the Mac's single 256 GB pool. Setup happens from a phone and a browser with LarePass, and ComfyUI and Ollama are one click each in the Market. Then the race: FLUX 3 Klein, Qwen-Image, ten portrait close-ups, two Wan 2.2 video tests, MiniMax H3 video and H3 Image, plus a Blender render. We also point an AI agent like Claude Code at the box with the Olares command-line skills and the Olares Router.
The Olares won six of the seven AI tests and tied the seventh, FLUX 3, and it was up to 3.7 times faster on images and up to 1.5 times faster on video. What surprised me most was Wan's 14-billion-parameter model. It doesn't fit in the Olares' 24 GB, so the box has to shuffle it in and out of system memory, and it still finished first, in under 11 minutes against the Mac's 14. If you keep the biggest models loaded at full precision, the Mac's 256 GB is where it pulls ahead. For most image and video work, though, the cheaper box beat the one that costs more than twice as much.
If you make images and video with open models and you're tired of paying per generation, or you'd rather keep client work off the cloud, this is a box you can put on a shelf and drive from your laptop, even from a hotel room over the LarePass VPN. If you already work on a Mac, it adds NVIDIA speed without replacing your main computer, and you can add storage later for far less than Apple charges up front.
CHAPTERS
00:00 Olares One vs Mac Studio
00:24 Before you buy a Mac Studio
01:00 What we're testing
01:40 Unboxing
02:22 What's inside
03:30 Two kinds of memory
04:19 Setup from a Mac
05:47 First generations
06:34 The race
09:56 Price and power
10:35 Upgrades
11:52 Your agent, your Olares
12:58 3D rendering
13:17 Which one should you buy?
If you set one up, switch it to Performance Mode first, then tell me what you're running on it in the comments.
🖥️ Get the Olares One $600 off with our link (Black Friday Deal) -tinyurl.com/olares600
Save $300 after black Friday
💙⚡Become a member – tinyurl.com/blmember
If you're thinking about buying a Mac Studio for AI, watch this first. Olares sponsored this video and sent us an Olares One, a small personal AI computer with NVIDIA's laptop RTX 5090 and 96 GB of memory, so we raced it against the $11,299 Mac Studio M5 Ultra we just bought. Through our link the Olares One is $600 off, which brings it to $4,399. By the end you'll know which one is faster for local image and video models, what each one costs to buy and to upgrade, and which kind of work suits each.
We open it up to show the RTX 5090 laptop chip, the 24-core Intel Core Ultra 9, the two memory sticks and the free SSD slot, and explain why its 24 GB of graphics memory and 96 GB of system memory behave so differently from the Mac's single 256 GB pool. Setup happens from a phone and a browser with LarePass, and ComfyUI and Ollama are one click each in the Market. Then the race: FLUX 3 Klein, Qwen-Image, ten portrait close-ups, two Wan 2.2 video tests, MiniMax H3 video and H3 Image, plus a Blender render. We also point an AI agent like Claude Code at the box with the Olares command-line skills and the Olares Router.
The Olares won six of the seven AI tests and tied the seventh, FLUX 3, and it was up to 3.7 times faster on images and up to 1.5 times faster on video. What surprised me most was Wan's 14-billion-parameter model. It doesn't fit in the Olares' 24 GB, so the box has to shuffle it in and out of system memory, and it still finished first, in under 11 minutes against the Mac's 14. If you keep the biggest models loaded at full precision, the Mac's 256 GB is where it pulls ahead. For most image and video work, though, the cheaper box beat the one that costs more than twice as much.
If you make images and video with open models and you're tired of paying per generation, or you'd rather keep client work off the cloud, this is a box you can put on a shelf and drive from your laptop, even from a hotel room over the LarePass VPN. If you already work on a Mac, it adds NVIDIA speed without replacing your main computer, and you can add storage later for far less than Apple charges up front.
CHAPTERS
00:00 Olares One vs Mac Studio
00:24 Before you buy a Mac Studio
01:00 What we're testing
01:40 Unboxing
02:22 What's inside
03:30 Two kinds of memory
04:19 Setup from a Mac
05:47 First generations
06:34 The race
09:56 Price and power
10:35 Upgrades
11:52 Your agent, your Olares
12:58 3D rendering
13:17 Which one should you buy?
If you set one up, switch it to Performance Mode first, then tell me what you're running on it in the comments.










