GoKawiilDeltafin fork runs full 2.8-trillion-parameter Kimi K3 model from SSDs on a MacBook Pro
A project called ARGODRIVE Deltafin, forked from gavamedia/deltafin, streams the complete, unpruned 2.8-trillion-parameter Kimi K3 model off four SSDs to run on an Apple Silicon M1 Max laptop. The latest benchmark shows throughput of about 0.29 tokens per second, roughly 2% faster than the prior measurement, with earlier logs showing the speed climbing from 0.0141 tokens/s in late July to current levels through iterative optimization.
GoKawiil's interpretation of the reporting above, not reported fact.
Running a model this large on consumer hardware, even at sub-1-token-per-second speeds, demonstrates that massive frontier-scale models can be executed without cloud infrastructure or quantization tricks that degrade output quality. The project's insistence on keeping all 16 experts active and never pruning weights makes it a useful benchmark for how far storage-bound inference can be pushed on ordinary laptops, even if practical usability remains far off.
- Deltafin streams Kimi K3's full 2.8T parameters from four SSDs rather than loading them into memory.
- Current speed on an M1 Max MacBook Pro is about 0.29 tokens/sec, a steady improvement from earlier tests.
- The project explicitly avoids quality-reducing shortcuts like pruning, treating this as a hardware-limits experiment rather than a usable product.
Samsung T7 Shield Portable SSD — Streaming a 2.8-trillion-parameter model off disk lives or dies on SSD throughput, and the T7 Shield is a rugged, bus-powered USB-C drive that plugs straight into a MacBook Pro. Grab several and you can mirror the multi-drive setup the Deltafin benchmark uses — or just enjoy fast external storage for huge model weights.
See Samsung T7 Shield Portable SSD on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: github.com, 2026-09-08
Published there as: “Kimi K3 (2.8T) at 1 token/s on a MacBook Pro, streamed from four SSDs”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.