Builds / Apple M3 Ultra (80-core GPU, unified)
Mac Studio M3 Ultra 512GB (oMLX, MTP off): Inkling 2 Mini at 40.3 tok/s at 1k context, 28.4 tok/s at 195k
Last updated:
- GPU(s)
- Apple M3 Ultra (80-core GPU, unified)
- VRAM / unified memory (GB)
- 512
- CPU
- Apple M3 Ultra (80-core GPU)
- Motherboard
- not stated
- RAM
- 512 GB unified memory
- PSU
- not stated
- Other
- oMLX v0.8.1, macOS 27.2.1; model labelled Inkling-2-Mini; MTP off; turboquant KV off; benchmark context Code (Python); temperature 1
- Reported total cost
- not stated
- Power
- not stated
Measured speed (generation tokens/sec)
| Model | Quant | Engine | tok/s | Prompt processing (tok/s) | Details |
|---|---|---|---|---|---|
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 40.3 | 1,169 @ 1k-token prompt est. TTFT @8k โ 7.0 s (calculated from measured prefill) | context: 1k tokens๐ Public oMLX session on an M3 Ultra (80-core) with 512 GB unified memory. Peak mem 20.9 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 39.5 | 1,367 @ 4k-token prompt est. TTFT @8k โ 6.0 s (calculated from measured prefill) | context: 4k tokens๐ Same oMLX session as the 1k row. Peak mem 22.7 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 39.4 | 1,360 @ 8k-token prompt est. TTFT @8k โ 6.0 s (calculated from measured prefill) | context: 8k tokens๐ Same oMLX session as the 1k row. Peak mem 23.4 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 38.7 | 1,328 @ 16k-token prompt est. TTFT @8k โ 6.2 s (calculated from measured prefill) | context: 16k tokens๐ Same oMLX session as the 1k row. Peak mem 24.9 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 37.3 | 1,256 @ 32k-token prompt est. TTFT @8k โ 6.5 s (calculated from measured prefill) | context: 32k tokens๐ Same oMLX session as the 1k row. Peak mem 27.9 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 34.4 | 1,130 @ 64k-token prompt est. TTFT @8k โ 7.2 s (calculated from measured prefill) measured TTFT: 58.0 s @ 65,536-token prompt | context: 64k tokens๐ Same oMLX session as the 1k row. Peak mem 33.8 GB. TTFT 57,976 ms as published on the 64k result page. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 30.9 | 941 @ 128k-token prompt est. TTFT @8k โ 8.7 s (calculated from measured prefill) | context: 128k tokens๐ Same oMLX session as the 1k row. Peak mem 45.8 GB. Retrieved 21 Nov 2026. |
| Inkling 2 Mini | oMLX Inkling-2-Mini (as labelled) | oMLX v0.8.1, MTP off | 28.4 | 801 @ 195k-token prompt est. TTFT @8k โ 10.2 s (calculated from measured prefill) | context: 195k tokens๐ Same oMLX session as the 1k row. Peak mem 58.5 GB. Retrieved 21 Nov 2026. |
Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 รท prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.
Public oMLX benchmark session, retrieved 21 Nov 2026. Hardware, software, settings and the context-length table are taken from the session page. MTP and turboquant KV were off. RigBase did not re-run it. Session resource panel: peak footprint 67.15 GB, total RAM 512.00 GB.
Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.
Quick answers: Can Apple M3 Ultra run Inkling 2 Mini?