Mac Studio (M5 Ultra, 2026)
M5
Apple's fastest computer: up to a 36-core CPU, an 80-core GPU and 1.2TB/s of memory bandwidth, with 96GB standard and up to 512GB. That memory is the point - 256GB or 512GB holds open models with hundreds of billions of parameters on one desk, and several can be clustered over Thunderbolt 5. For anything that fits in 128GB the M5 Max costs far less.
Against the line-up
Geekbench 6 CPU
Geekbench 6 CPU
Geekbench 6 GPU
Blender Benchmark
The whole Mac Studio line-up side by side, with prices
Every benchmark
Quoted from the publications that ran them; we don't test hardware ourselves. Figures from different publications were run on different systems and scenes, so compare within a row's source.
Local LLM (llama.cpp / Ollama)
| Setting | Result | Source |
|---|---|---|
| Llama 2 7B F16, Generation |
73.6tok/s | llama.cpp community benchmarks (GitHub), Oct 19, 2026 M5 Ultra, 80-core GPU, 1228GB/s (configure-to-order) llama-bench tg128 (generating 128 tokens), all layers on the GPU. |
| Llama 2 7B Q4_0, Generation |
179tok/s | llama.cpp community benchmarks (GitHub), Oct 19, 2026 M5 Ultra, 80-core GPU, 1228GB/s (configure-to-order) llama-bench tg128 (generating 128 tokens), all layers on the GPU. No 64-core base-model figure yet. |
| Llama 2 7B Q4_0, Prompt processing |
4,945tok/s | llama.cpp community benchmarks (GitHub), Oct 19, 2026 M5 Ultra, 80-core GPU, 1228GB/s (configure-to-order) llama-bench pp512 (reading a 512-token prompt), all layers on the GPU. |
Blender Benchmark
| Setting | Result | Source |
|---|---|---|
| Standard run |
13,502samples/min | Blender Open Data, Oct 19, 2026 Apple M5 Ultra (GPU - 64 cores), Metal, Blender 5.2.0 Median score over the three scenes, 7 submitted runs |
| 80-core GPU |
15,327samples/min | Blender Open Data, Oct 19, 2026 Apple M5 Ultra (GPU - 80 cores), Metal, Blender 5.2.0 Median score over the three scenes, 6 submitted runs |
| CPU |
1,142samples/min | Blender Open Data, Oct 19, 2026 Apple M5 Ultra CPU, Blender 5.2.0 Median score over the three scenes, 14 submitted runs. Blender Open Data does not split the 30- and 36-core CPUs. |
Geekbench 6 CPU
| Setting | Result | Source |
|---|---|---|
| Multi-core |
37,870points | The Verge, Sep 21, 2026 Mac Studio M5 Ultra, CTO 36-core CPU, 80-core GPU, 256GB, 4TB |
| Single-core |
4,310points | The Verge, Sep 21, 2026 Mac Studio M5 Ultra, CTO 36-core CPU, 80-core GPU, 256GB, 4TB |
Geekbench 6 GPU
Sources
- Blender Open Data (results database)
- The Verge (review)
- llama.cpp community benchmarks (GitHub) (results database)
As an Amazon Associate we earn from qualifying purchases, and we may earn from other shop links. Amazon's price is on Amazon; Best Buy's is read from Best Buy and shown with the time it was checked.