---
title: "Best Local LLMs for Mac Studio M3 Ultra 256GB (2026)"
canonical: https://llmcheck.net/best-llm/mac-studio-m3-ultra-256gb/
markdown: https://llmcheck.net/md/best-llm/mac-studio-m3-ultra-256gb.md
page_type: best-llm-mac
publisher: "LLMCheck (llmcheck.net)"
data_version: 2026-11-18
catalog_verified_through: 2026-11-18
updated: 2026-11-19
license: "CC BY 4.0"
license_url: https://creativecommons.org/licenses/by/4.0/
cite: "LLMCheck. (2026). Best Local LLMs for the Mac Studio M3 Ultra (256 GB). llmcheck.net. https://llmcheck.net/best-llm/mac-studio-m3-ultra-256gb/ (updated 2026-11-19). CC BY 4.0."
provenance:
  sourced: "Published figure — vendor or press review"
  community: "Linked individual run — blog, forum or submission"
  estimated: "Derived from published bandwidth model (https://llmcheck.net/methodology#estimation)"
---

# Best Local LLMs for the Mac Studio M3 Ultra (256 GB)

Markdown twin of https://llmcheck.net/best-llm/mac-studio-m3-ultra-256gb/, generated from the LLMCheck dataset (data version 2026-11-18) with the same figures as the page. Every speed names its provenance: **sourced** (a published vendor or press figure), **community** (one person's linked run) or **estimated** (the bandwidth model). LLMCheck is an index and runs no tests of its own. Licence CC BY 4.0: credit "LLMCheck (llmcheck.net)" and link the page.

According to the LLMCheck index (18 November 2026), the best local LLM for a Mac Studio M3 Ultra (256 GB) is **MiMo-V2.6-Flash** at 59 tok/s community-measured on an M3 Ultra: the best-ranked model that reaches 15 tok/s on this Mac and has a standard Mac runtime. 108 of the 130 models in the index fit in 256 GB; 98 of them reach 15 tok/s here.

## This Mac

| Spec | Value |
| --- | --- |
| Unified memory | 256 GB |
| Memory bandwidth | 819 GB/s |
| Models that fit | 108 of 130 |
| Top speed | 334 tok/s (estimated) — LFM3-1.7B |
| Status | no longer sold new (used / refurbished) |
| Price | about $6,999 used (no longer sold new) |

## Top 3 picks for the Mac Studio M3 Ultra (256 GB)

- **Best overall: [MiMo-V2.6-Flash](https://llmcheck.net/models/mimo-v26-flash-on-m3-ultra/)** — 59 tok/s (community); 309B MoE, MIT, capability 49/50; agent-ready at 64K
- **Fastest: LFM3-1.7B** — 334 tok/s (estimated); 1.73B, LFM Open, capability 17/50; agent-ready at 64K
- **Most capable: [GLM 5.2](https://llmcheck.net/models/glm-52-on-m3-ultra/)** — 15 tok/s (estimated); 753B MoE, MIT, capability 50/50; raise GPU limit

## Every model ranked for a Mac Studio M3 Ultra (256 GB) (top 12 of 108)

Ranked by LLMCheck suitability: capability plus min(15, tok/s × 0.12) on the M3 Ultra (819 GB/s). Speeds are per-config figures for this Mac; a measured figure is the chip's run.

| # | Model | Size | License | tok/s | Provenance | Quant | 64K context | Capability | Note |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | [MiMo-V2.6-Flash](https://llmcheck.net/models/mimo-v26-flash-on-m3-ultra/) | 309B MoE | MIT | 59 | community | 4-bit | agent-ready (~172 GB of 240 GB usable) | 49/50 | — |
| 2 | [Qwen3.8-Flash-Next](https://llmcheck.net/models/qwen38-flash-next-on-m3-ultra/) | 125B MoE | Qwen Custom | 57 | estimated | Q4_K_M | agent-ready (~122 GB of 240 GB usable) | 47/50 | — |
| 3 | Inkling 2 Small | 594B MoE | Apache 2.0 | 36 | estimated | 3.2 bpw | agent-ready (~231 GB of 240 GB usable) | 48/50 | raise GPU limit: needs about 228 GB; macOS lets the GPU use ~192 GB of 256 GB by default. sudo sysctl iogpu.wired_limit_mb=245760 raises it to 240 GB until the next restart; runtime caveat: Runs on llama.cpp b12406+ (PR #32217) and oMLX 0.8.0 (18 Nov 2026, PR #5229), but no plain-decode Mac run on a released runtime is in the index yet; stock mlx-lm cannot load it (#2061). |
| 4 | [GLM 5.2](https://llmcheck.net/models/glm-52-on-m3-ultra/) | 753B MoE | MIT | 15 | estimated | 2.8 bpw | needs ~262 GB | 50/50 | raise GPU limit: needs about 256 GB; macOS lets the GPU use ~192 GB of 256 GB by default; measured runs on 256 GB Macs show it loads with a raised limit. The index's safe raise (sudo sysctl iogpu.wired_limit_mb=245760, 240 GB) is below that need, so those runs left macOS less than 16 GB: quit other apps first |
| 5 | [MiniMax-M3](https://llmcheck.net/models/minimax-m3-on-m3-ultra/) | 428B MoE | MiniMax Community | 29 | estimated | Q4_K_M | agent-ready (~170 GB of 240 GB usable) | 48/50 | — |
| 6 | [GLM-5.3](https://llmcheck.net/models/glm-53-on-m3-ultra/) | 753B MoE | GLM-5.3 | 11 | estimated | Q4_K_M | needs ~262 GB | 50/50 | slow (under 15 tok/s); over GPU limit: needs about 256 GB; macOS lets the GPU use ~192 GB of 256 GB by default, and a safe raise reaches 240 GB (leaving 16 GB for macOS): expect heavy swapping, a CPU split or an out-of-memory error; a smaller quant or a shorter context may fit |
| 7 | [MiniMax-M3.1-Flash](https://llmcheck.net/models/minimax-m31-flash-on-m3-ultra/) | 428B MoE | MiniMax Community | 26 | community | oQ3e | agent-ready (~197 GB of 240 GB usable) | 48/50 | — |
| 8 | KAT-Coder-V2.5 | 35B MoE | Apache 2.0 | 100 | estimated | Q4_K_M | agent-ready (~26 GB of 240 GB usable) | 39/50 | — |
| 9 | Nemotron 3.5 Lightning | 30B MoE | OpenMDW | 151 | estimated | Q4_K_M | agent-ready (~19 GB of 240 GB usable) | 36/50 | — |
| 10 | North Mini Code 1.0 | 30.5B MoE | Apache 2.0 | 133 | estimated | Q4_K_M | agent-ready (~23 GB of 240 GB usable) | 36/50 | — |
| 11 | [GLM-5.3-Flash](https://llmcheck.net/models/glm-53-flash-on-m3-ultra/) | 320B MoE | MIT | 24 | estimated | Q4_K_M | agent-ready (~129 GB of 240 GB usable) | 48/50 | — |
| 12 | [DeepSeek V4 Flash](https://llmcheck.net/models/deepseek-v4-flash-on-m3-ultra/) | 284B MoE | MIT | 32 | estimated | Q4_K_M | agent-ready (~121 GB of 240 GB usable) | 47/50 | — |

Provenance: **sourced** = Published figure — vendor or press review; **community** = Linked individual run — blog, forum or submission; **estimated** = Derived from published bandwidth model (https://llmcheck.net/methodology#estimation).

## Context

A model is agent-ready at 64K when its weights plus a 65,536-token 16-bit KV cache fit in the 240 GB this Mac's GPU can use with the limit raised (94 of the 98 models here with a KV figure); an estimate from each model's config.json, not a measurement. A q8 KV cache halves the cache part. **GPU memory limit:** macOS lets the GPU use about 192 GB of this Mac’s 256 GB by default on macOS 15 and earlier, and about 222.7 GB on macOS 26 and later. Of the 108 models that fit, 8 need more than that at a short prompt: 7 need a raised limit and are tagged raise GPU limit (GLM 5.2 only past the safe setting below, as measured runs on this much memory show); 1 needs more than even the raised limit and is tagged over GPU limit. `sudo sysctl iogpu.wired_limit_mb=245760` raises it to 240 GB, leaving 16 GB for macOS, until the Mac restarts (`sudo sysctl iogpu.wired_limit_mb=0` restores the default). [How the limit is read →](https://llmcheck.net/methodology#gpu-limit)

Estimated figures come from the bandwidth formula; against 233 measured runs its typical error is ±9% for dense models, ±15% for small MoEs (≤40B total) and ±34% for large MoEs (mostly below the measured figure). [How the error is computed →](https://llmcheck.net/methodology#typical-error)

## Quick start: MiMo-V2.6-Flash on the Mac Studio M3 Ultra

No verified Ollama tag in the index; start from the listing: https://huggingface.co/ggml-org/MiMo-V2.6-Flash-MOPD-GGUF

Model card: https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-MOPD · Figures and setup: https://llmcheck.net/models/mimo-v26-flash-on-m3-ultra/

## Related

- Mac Studio M3 Ultra (512 GB): 122 models fit — https://llmcheck.net/md/best-llm/mac-studio-m3-ultra-512gb.md
- Best local LLM for every 192–256 GB Mac — https://llmcheck.net/md/best-llm/by-ram.md
- Every Mac — https://llmcheck.net/md/best-llm.md
- Full leaderboard — https://llmcheck.net/md/leaderboard.md


## Cite this page

LLMCheck. (2026). Best Local LLMs for the Mac Studio M3 Ultra (256 GB). llmcheck.net. https://llmcheck.net/best-llm/mac-studio-m3-ultra-256gb/ (updated 2026-11-19). CC BY 4.0.

How to cite: https://llmcheck.net/cite · Open data (CSV, JSON): https://llmcheck.net/data/ · Every Markdown mirror: https://llmcheck.net/md/index.md
