--- title: "#apple-silicon articles | Hashnode" description: "Explore 40 articles tagged #apple-silicon on Hashnode. Tutorials, guides, and how-tos from the community." url: "https://hashnode.com/tag/apple-silicon" --- Tag # #apple-silicon 40 posts · 6 followers [Robert Collins](https://hashnode.com/@robbiemu) · [5d ago](https://hashnode.com/posts/from-numerical-differences-to-bitwise-prediction-in-core-ml-arithmetic/6aee002668becd2cfcbf0974) ### [From Numerical Differences to Bitwise Prediction in Core ML Arithmetic](https://selfenrichment.hashnode.dev/from-numerical-differences-to-bitwise-prediction-in-core-ml-arithmetic?utm_source=hashnode&utm_medium=feed) How targeted probes recover Core ML arithmetic rules and predict outputs bit for bit across tested CPU, GPU, and Neural Engine paths. 77 min read · [selfenrichment.hashnode.dev](https://selfenrichment.hashnode.dev/from-numerical-differences-to-bitwise-prediction-in-core-ml-arithmetic?utm_source=hashnode&utm_medium=feed) [Mihai\_LeanZero](https://hashnode.com/@mihaileanzero) · [Sep 19](https://hashnode.com/posts/mlx-vs-gguf-on-apple-silicon-benchmarking-the-same-local-model-two-ways/6aaeb06506ef87fd984a3ef5) ### [MLX vs GGUF on Apple Silicon: Benchmarking the Same Local Model Two Ways](https://leanzero.hashnode.dev/mlx-vs-gguf-on-apple-silicon-benchmarking-the-same-local-model-two-ways?utm_source=hashnode&utm_medium=feed) One 27B model as GGUF and MLX, 30 app builds graded by running them. Each worked 13 of 15, speed was a wash, and MLX was steadier: the safe default. 13 min read · [leanzero.hashnode.dev](https://leanzero.hashnode.dev/mlx-vs-gguf-on-apple-silicon-benchmarking-the-same-local-model-two-ways?utm_source=hashnode&utm_medium=feed) [Mihai\_LeanZero](https://hashnode.com/@mihaileanzero) · [Sep 19](https://hashnode.com/posts/ling-30-flash-wont-load-on-a-mac-studio-qwen3-coder-next-does-at-73-toks/6aae2fb6e1223f8c1c86b31b) ### [Ling 3.0 Flash won't load on a Mac Studio. Qwen3-Coder-Next does, at 73 tok/s.](https://leanzero.hashnode.dev/ling-30-flash-wont-load-on-a-mac-studio-qwen3-coder-next-does-at-73-toks?utm_source=hashnode&utm_medium=feed) Ant Group's Ling 3.0 Flash still won't load in mlx-lm on a 96 GB Mac Studio: 385 of 62,237 tensors block it. Qwen3-Coder-Next runs, at 73 tok/s. 23 min read · [leanzero.hashnode.dev](https://leanzero.hashnode.dev/ling-30-flash-wont-load-on-a-mac-studio-qwen3-coder-next-does-at-73-toks?utm_source=hashnode&utm_medium=feed) [Khasky](https://hashnode.com/@khasky) · [Sep 12](https://hashnode.com/posts/edge0-streams-moe-experts-off-ssd-to-fit-35b-in-3-gb/6aa4e2662d1f4c30368b708c) ### [Edge0 Streams MoE Experts off SSD to Fit 35B in 3 GB](https://khasky.hashnode.dev/edge0-streams-moe-experts-off-ssd-to-fit-35b-in-3-gb?utm_source=hashnode&utm_medium=feed) One question took me through Edge0: if the 35B model only needs 2.9 GB of RAM, what does the machine actually need? The answer is a disk budget, a cache budget and a context budget, and none of them i 5 min read · [khasky.hashnode.dev](https://khasky.hashnode.dev/edge0-streams-moe-experts-off-ssd-to-fit-35b-in-3-gb?utm_source=hashnode&utm_medium=feed) [Billie M](https://hashnode.com/@billiem) · [Sep 9](https://hashnode.com/posts/qwen3-8-27b-on-m4-pro/6aa1c2f1e9d978acd85efffb) ### [Local models are actually good now - playing with Qwen3.8-27B](https://billiem.hashnode.dev/qwen3-8-27b-on-m4-pro?utm_source=hashnode&utm_medium=feed) A matched Pi experiment shows why quantisation is an agent memory decision: 4-bit and 6-bit finished, while 8-bit hit Metal OOM. 7 min read · [billiem.hashnode.dev](https://billiem.hashnode.dev/qwen3-8-27b-on-m4-pro?utm_source=hashnode&utm_medium=feed) [Oliver January](https://hashnode.com/@neutrontech-ai) · [Aug 31](https://hashnode.com/posts/is-cloud-indexing-necessary-a-guide-to-local-first-workspace-software/6a95c54622c64fc15a7d7691) ### [Is Cloud Indexing Necessary? A Guide to Local-First Workspace Software](https://neutrontech.hashnode.dev/is-cloud-indexing-necessary-a-guide-to-local-first-workspace-software?utm_source=hashnode&utm_medium=feed) Does workspace really need cloud indexing? Explore cloud vs. local indexing, on-device AI, data sovereignty, and the rise of local-first workspaces. 12 min read · [neutrontech.hashnode.dev](https://neutrontech.hashnode.dev/is-cloud-indexing-necessary-a-guide-to-local-first-workspace-software?utm_source=hashnode&utm_medium=feed) [Manu Shukla](https://hashnode.com/@mr-manushukla) · [Aug 3](https://hashnode.com/posts/apple-silicon-vs-cloud-apis-in-2026-the-local-llm-break-even-with-real-benchmarks/6a70e44cb263fead21e5e954) ### [Apple Silicon vs cloud APIs in 2026: the local LLM break-even, with real benchmarks](https://ecorpit.hashnode.dev/apple-silicon-vs-cloud-apis-in-2026-the-local-llm-break-even-with-real-benchmarks?utm_source=hashnode&utm_medium=feed) Apple Silicon vs cloud APIs in 2026: the local LLM break-even, with real benchmarks Summary. As of 4 August 2026 Apple's Mac Studio specification page lists exactly 2 chips: M4 Max at 410GB/s of memor 15 min read · [ecorpit.hashnode.dev](https://ecorpit.hashnode.dev/apple-silicon-vs-cloud-apis-in-2026-the-local-llm-break-even-with-real-benchmarks?utm_source=hashnode&utm_medium=feed) [1](https://hashnode.com/posts/apple-silicon-vs-cloud-apis-in-2026-the-local-llm-break-even-with-real-benchmarks/6a70e44cb263fead21e5e954) [Manu Shukla](https://hashnode.com/@mr-manushukla) · [Jul 18](https://hashnode.com/posts/pytorch-213-on-apple-silicon-flexattention-is-up-to-12x-faster-than-sdpa/6a5baeef299d84cdace2a7dd) ### [PyTorch 2.13 on Apple Silicon: FlexAttention is up to 12x faster than SDPA](https://ecorpit.hashnode.dev/pytorch-213-on-apple-silicon-flexattention-is-up-to-12x-faster-than-sdpa?utm_source=hashnode&utm_medium=feed) PyTorch 2.13 on Apple Silicon: FlexAttention is up to 12x faster than SDPA Summary. PyTorch 2.13, released around July 11, 2026, brings FlexAttention to Apple Silicon through the Metal (MPS) backend, 13 min read · [ecorpit.hashnode.dev](https://ecorpit.hashnode.dev/pytorch-213-on-apple-silicon-flexattention-is-up-to-12x-faster-than-sdpa?utm_source=hashnode&utm_medium=feed) [Manu Shukla](https://hashnode.com/@mr-manushukla) · [Jul 18](https://hashnode.com/posts/pytorch-213-brings-flexattention-to-apple-silicon-a-12x-sparse-attention-speedup/6a5b2362bb77b2f6fd2a493b) ### [PyTorch 2.13 brings FlexAttention to Apple Silicon: a 12x sparse-attention speedup](https://ecorpit.hashnode.dev/pytorch-213-brings-flexattention-to-apple-silicon-a-12x-sparse-attention-speedup?utm_source=hashnode&utm_medium=feed) PyTorch 2.13 brings FlexAttention to Apple Silicon: a 12x sparse-attention speedup Summary. PyTorch 2.13 shipped on July 8, 2026, and for anyone training or running models on a Mac the headline is one 12 min read · [ecorpit.hashnode.dev](https://ecorpit.hashnode.dev/pytorch-213-brings-flexattention-to-apple-silicon-a-12x-sparse-attention-speedup?utm_source=hashnode&utm_medium=feed) [Tina L](https://hashnode.com/@tiagol) · [Jul 5](https://hashnode.com/posts/microsoft-store-stuck-on-loading-in-windows-11-on-parallels-m1-mac/6a4a3b134be5528ed5dfac51) ### [Microsoft Store Stuck on Loading in Windows 11 on Parallels (M1 Mac)](https://debugnote.hashnode.dev/microsoft-store-stuck-on-loading-in-windows-11-on-parallels-m1-mac?utm_source=hashnode&utm_medium=feed) Microsoft Store stuck on loading in Windows 11 ARM on Parallels Desktop on M1 Mac. 2 min read · [debugnote.hashnode.dev](https://debugnote.hashnode.dev/microsoft-store-stuck-on-loading-in-windows-11-on-parallels-m1-mac?utm_source=hashnode&utm_medium=feed)