---
title: "Local LLM Leaderboard for Mac: 130 Open Models Ranked (2026)"
canonical: https://llmcheck.net/leaderboard
markdown: https://llmcheck.net/md/leaderboard.md
page_type: leaderboard
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). The local LLM leaderboard. llmcheck.net. https://llmcheck.net/leaderboard (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)"
---

# The local LLM leaderboard

Markdown twin of https://llmcheck.net/leaderboard, 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), MiMo-V2.6-Flash and Qwen 3.6-35B-A3B tie for #1 of 130 open-weight models at Score 81 (MiMo-V2.6-Flash is listed first on the alphabetical tie-break). MiMo-V2.6-Flash needs a Mac Studio with 192 GB or more; Qwen 3.6-35B-A3B needs 24 GB, the best pick at 32 GB or less. Every speed carries a provenance mark.

## All 130 models, ranked by LLMCheck Score

Speed = single-stream decode (tok/s) on the reference chip: the M5 Max (614 GB/s), or the M5 Ultra (1,228 GB/s) for models that need more than 128 GB; the lower-median measured plain-decode run at 5 bits or fewer on that chip where one exists, otherwise the estimate. The Score uses exactly the figure shown. RAM @64K = minimum RAM plus the 16-bit KV cache the rest of a 65,536-token context adds, computed from each model's config.json (an estimate, not a measurement). Prompt figures appear only where a run measured them.

| # | Model | Score | tok/s | Provenance | Chip | Quant | Min RAM | RAM @64K | License | Size | Company | Prompt (measured runs only) | Ollama tag |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | [MiMo-V2.6-Flash](https://llmcheck.net/models/mimo-v26-flash-on-m5-ultra/) | 81 | 77 | community | M5 Ultra | MXFP4 | 170 GB | 172 GB (estimated; 171 GB with a q8 KV cache) | MIT | 309B MoE | Xiaomi | prefill 2,755 tok/s @8K (community, M5 Ultra); TTFT 0.609 s @1K (community) | — |
| 2 | [Qwen 3.6-35B-A3B](https://llmcheck.net/models/qwen-36-35b-a3b-on-m5-max/) | 81 | 130 | community | M5 Max | 4-bit | 24 GB | 26 GB (estimated; 25 GB with a q8 KV cache) | Apache 2.0 | 35B MoE | Alibaba | prefill 4,576 tok/s @1K (community, M5 Max); TTFT 0.224 s @1K (community) | `qwen3.6:35b-a3b` |
| 3 | [GLM-4.7-Flash](https://llmcheck.net/models/glm-47-flash-on-m4-max/) | 80 | 107 | estimated | M5 Max | Q4_K_M | 18 GB | 22 GB (estimated; 20 GB with a q8 KV cache) | MIT | 31B MoE | Zhipu AI | — | `glm-4.7-flash` |
| 4 | [Recursive Flint](https://llmcheck.net/models/recursive-flint-on-m5-max/) | 80 | 123 | community | M5 Max | 4-bit | 20 GB | 22 GB (estimated; 21 GB with a q8 KV cache) | Apache 2.0 | 39.8B MoE | Recursive | prefill 3,592 tok/s @8K (community, M5 Max); TTFT 0.309 s @1K (community) | `flint:40b` |
| 5 | [KAT-Coder-V2.5](https://llmcheck.net/models/kat-coder-v25-on-m5-max/) | 79 | 89 | community | M5 Max | OptiQ-4bit | 24 GB | 26 GB (estimated; 25 GB with a q8 KV cache) | Apache 2.0 | 35B MoE | Kwaipilot | prefill 4,018 tok/s @4K (community, M5 Max); TTFT 0.316 s @1K (community) | — |
| 6 | [Nemotron 3.5 Lightning](https://llmcheck.net/models/nemotron-35-lightning-on-m5-max/) | 79 | 117 | community | M5 Max | Q4_0 | 18 GB | 19 GB (estimated; 19 GB with a q8 KV cache) | OpenMDW | 30B MoE | NVIDIA | — | `nemotron-3.5-lightning:30b` |
| 7 | [North Mini Code 1.0](https://llmcheck.net/models/north-mini-code-10-on-m5-max/) | 79 | 120 | community | M5 Max | 4-bit | 21 GB | 23 GB (estimated; 22 GB with a q8 KV cache) | Apache 2.0 | 30.5B MoE | Cohere | prefill 3,424 tok/s @8K (community, M5 Max); TTFT 0.376 s @1K (community) | `north-mini-code-1.0` |
| 8 | [Gemma 4 26B-A4B](https://llmcheck.net/models/gemma-4-26b-a4b-on-m5-max/) | 78 | 107 | sourced | M5 Max | Q4_K_M | 18 GB | 20 GB (estimated; 19 GB with a q8 KV cache) | Apache 2.0 | 26B MoE | Google | prefill 3,852 tok/s @4K (community, M5 Max) | `gemma4:26b-a4b` |
| 9 | [MiniCPM-V 4.7 35B-A3B](https://llmcheck.net/models/minicpm-v-47-35b-a3b-on-m5-max/) | 78 | 100 | community | M5 Max | MXFP4 | 23 GB | 25 GB (estimated; 24 GB with a q8 KV cache) | Apache 2.0 | 35.2B MoE | OpenBMB | prefill 3,583 tok/s @4K (community, M5 Max); TTFT 0.804 s @1K (community) | `openbmb/minicpm-v4.7:35b-a3b` |
| 10 | [Raon-1 mini preview](https://llmcheck.net/models/raon-1-mini-preview-on-m5-max/) | 77 | 123 | estimated | M5 Max | Q4_K_M | 18 GB | 20 GB (estimated; 19 GB with a q8 KV cache) | Apache 2.0 | 28.8B MoE | KRAFTON | — | — |
| 11 | [LFM2.5-24B-A2B](https://llmcheck.net/models/lfm25-24b-a2b-on-m5-max/) | 76 | 184 | sourced | M5 Max | Q4_K_M | 17 GB | 19 GB (estimated; 18 GB with a q8 KV cache) | LFM Open | 24B MoE | Liquid AI | prefill 3,772 tok/s @4K (community, M5 Max) | `lfm2.5:24b` |
| 12 | Laguna XS 2.1 | 76 | 117 | estimated | M5 Max | Q4_K_M | 24 GB | 27 GB (estimated; 26 GB with a q8 KV cache) | OpenMDW | 33B MoE | Poolside | — | — |
| 13 | [Bonsai 27B](https://llmcheck.net/models/bonsai-27b-on-m5-max/) | 72 | 66 | sourced | M5 Max | Q1_0_g128 | 6 GB | 11 GB (estimated; 9 GB with a q8 KV cache) | Apache 2.0 | 27B | Prism ML | — | — |
| 14 | [Qwen3.8-27B](https://llmcheck.net/models/qwen38-27b-on-m5-max/) | 72 | 31 | community | M5 Max | 4-bit | 19 GB | 24 GB (estimated; 22 GB with a q8 KV cache) | Apache 2.0 | 27.8B | Alibaba | prefill 839 tok/s @4K (community, M5 Max) | `qwen3.8:27b` |
| 15 | Nemotron-Cascade 2 | 71 | 117 | estimated | M5 Max | Q4_K_M | 18 GB | 19 GB (estimated; 19 GB with a q8 KV cache) | NVIDIA Open | 30B MoE | NVIDIA | — | `nemotron-cascade-2:30b` |
| 16 | [DeepSeek V4 Flash](https://llmcheck.net/models/deepseek-v4-flash-on-m5-max/) | 70 | 39 | community | M5 Max | Q2_K | 120 GB | 121 GB (estimated; 121 GB with a q8 KV cache) | MIT | 284B MoE | DeepSeek | TTFT 0.4 s (community) | — |
| 17 | [DeepSeek V4 Flash Vision-Exp](https://llmcheck.net/models/deepseek-v4-flash-vision-exp-on-m5-max/) | 70 | 39 | estimated | M5 Max | Q2_K | 120 GB | 121 GB (estimated; 121 GB with a q8 KV cache) | MIT | 305B MoE | DeepSeek | — | — |
| 18 | [DeepSeek V4.1 Flash](https://llmcheck.net/models/deepseek-v41-flash-on-m5-ultra/) | 70 | 29 | estimated | M5 Ultra | oQ4e | 434 GB | 435 GB (estimated; 435 GB with a q8 KV cache) | MIT | 552B MoE | DeepSeek | — | — |
| 19 | [K2-Horizon-3.7B](https://llmcheck.net/models/k2-horizon-37b-on-m5-max/) | 70 | 133 | estimated | M5 Max | Q4_K_M | 4 GB | 14 GB (estimated; 9 GB with a q8 KV cache) | Apache 2.0 | 5.1B | MBZUAI IFM | — | — |
| 20 | [Ling-3.0-flash-VL](https://llmcheck.net/models/ling-30-flash-vl-on-m5-max/) | 70 | 53 | estimated | M5 Max | Q4_K_M | 56 GB | 57 GB (estimated; 57 GB with a q8 KV cache) | MIT | 124B MoE | Ant Group | — | — |
| 21 | [Maple Preview 20B-A1B](https://llmcheck.net/models/maple-preview-20b-a1b-on-m5-pro/) | 70 | 387 | sourced | M5 Pro | Ternary | 6 GB | 7 GB (estimated; 7 GB with a q8 KV cache) | MIT | 20B MoE | DeepGrove | — | — |
| 22 | [Nanbeige4.3-33B](https://llmcheck.net/models/nanbeige43-33b-on-m5-max/) | 70 | 25 | community | M5 Max | oQ4e | 17 GB | 31 GB (estimated; 24 GB with a q8 KV cache) | Apache 2.0 | 33.1B | Nanbeige | prefill 613 tok/s @4K (community, M5 Max); TTFT 1.665 s @1K (community) | — |
| 23 | [Nemotron 3.5 Bolt 74B-A9B](https://llmcheck.net/models/nemotron-35-bolt-74b-a9b-on-m5-max/) | 70 | 64 | community | M5 Max | 5-bit | 45 GB | 46 GB (estimated; 46 GB with a q8 KV cache) | OpenMDW | 74B MoE | NVIDIA | prefill 1,332 tok/s @4K (community, M5 Max) | — |
| 24 | [Qwen 3.5 35B-A3B](https://llmcheck.net/models/qwen-35-35b-a3b-on-m5-ultra/) | 70 | 136 | estimated | M5 Max | Q4_K_M | 24 GB | 26 GB (estimated; 25 GB with a q8 KV cache) | Apache 2.0 | 35B MoE | Alibaba | — | `qwen3.5:35b-a3b` |
| 25 | [Qwen 3.6-27B](https://llmcheck.net/models/qwen-36-27b-on-m5-max/) | 70 | 30 | estimated | M5 Max | Q4_K_M | 18 GB | 23 GB (estimated; 21 GB with a q8 KV cache) | Apache 2.0 | 27B | Alibaba | — | `qwen3.6:27b` |
| 26 | [Qwen3-Coder-Next](https://llmcheck.net/models/qwen3-coder-next-on-m5-max/) | 70 | 84 | estimated | M5 Max | Q4_K_M | 48 GB | 50 GB (estimated; 49 GB with a q8 KV cache) | Apache 2.0 | 80B MoE | Alibaba | — | `qwen3-coder-next` |
| 27 | [GLM 5.2](https://llmcheck.net/models/glm-52-on-m3-ultra/) | 69 | 22 | estimated | M5 Ultra | Q4_K_M | 256 GB | 262 GB (estimated; 259 GB with a q8 KV cache) | MIT | 753B MoE | Zhipu AI | — | — |
| 28 | [MiniCPM5-2B](https://llmcheck.net/models/minicpm5-2b-on-m5-max/) | 69 | 225 | estimated | M5 Max | Q4_K_M | 3 GB | 6 GB (estimated; 5 GB with a q8 KV cache) | Apache 2.0 | 2.52B | OpenBMB | — | — |
| 29 | [Ternary Bonsai 2 27B](https://llmcheck.net/models/ternary-bonsai-2-27b-on-m5-max/) | 69 | 47 | sourced | M5 Max | PQ2_0 | 9 GB | 14 GB (estimated; 12 GB with a q8 KV cache) | Apache 2.0 | 27.36B | Prism ML | — | `bonsai-2:27b` |
| 30 | [GLM-5.3-Flash](https://llmcheck.net/models/glm-53-flash-on-m5-max/) | 68 | 27 | estimated | M5 Max | IQ3_XXS | 128 GB | 129 GB (estimated; 129 GB with a q8 KV cache) | MIT | 320B MoE | Zhipu AI | — | — |
| 31 | [GPT-oss 120B](https://llmcheck.net/models/gpt-oss-120b-on-m5-max/) | 68 | 92 | community | M5 Max | MXFP4 | 62 GB | 65 GB (estimated; 64 GB with a q8 KV cache) | Apache 2.0 | 117B MoE | OpenAI | prefill 2,022 tok/s @4K (community, M5 Max); TTFT 1.109 s @1K (community) | `gpt-oss:120b` |
| 32 | [Inkling 2 Small](https://llmcheck.net/models/inkling-2-small-on-m5-ultra/) | 68 | 36 | estimated | M5 Ultra | Q4_K_M | 228 GB | 231 GB (estimated; 230 GB with a q8 KV cache) | Apache 2.0 | 594B MoE | Thinking Machines | — | — |
| 33 | [Kolibri-1](https://llmcheck.net/models/kolibri-1-on-m5-max/) | 68 | 75 | estimated | M5 Max | Q4_K_M | 50 GB | 52 GB (estimated; 51 GB with a q8 KV cache) | Apache 2.0 | 78.1B MoE | Aleph Alpha | — | — |
| 34 | [Ling-3.0 Flash](https://llmcheck.net/models/ling-30-flash-on-m5-max/) | 68 | 68 | community | M5 Max | 4-bit | 96 GB | 97 GB (estimated; 97 GB with a q8 KV cache) | MIT | 124B MoE | Ant Group | prefill 1,615 tok/s @4K (community, M5 Max); TTFT 0.644 s @1K (community) | — |
| 35 | Nanbeige4.2-3B | 68 | 199 | estimated | M5 Max | Q4_K_M | 2 GB | 8 GB (estimated; 5 GB with a q8 KV cache) | Apache 2.0 | 3B | Nanbeige | — | — |
| 36 | [Spark-X2.6-4B](https://llmcheck.net/models/spark-x26-4b-on-m5-max/) | 68 | 102 | community | M5 Max | oQ5e | 4 GB | 7 GB (estimated; 6 GB with a q8 KV cache) | Apache 2.0 | 4.11B | SparkLLM (XHToken) | prefill 2,867 tok/s @8K (community, M5 Max); TTFT 0.237 s @1K (community) | — |
| 37 | [GPT-oss 20B](https://llmcheck.net/models/gpt-oss-20b-on-m5-max/) | 67 | 141 | community | M5 Max | MXFP4 | 13 GB | 15 GB (estimated; 14 GB with a q8 KV cache) | Apache 2.0 | 21B MoE | OpenAI | prefill 4,774 tok/s @4K (community, M5 Max); TTFT 0.587 s @1K (community) | `gpt-oss:20b` |
| 38 | [Inkling-Small](https://llmcheck.net/models/inkling-small-on-m5-ultra/) | 67 | 42 | estimated | M5 Max | Q2_K | 96 GB | 98 GB (estimated; 97 GB with a q8 KV cache) | Apache 2.0 | 276B MoE | Thinking Machines | — | — |
| 39 | Kimi K2.6 | 67 | 27 | estimated | M5 Ultra | 3.6 bpw | 429 GB | 434 GB (estimated; 432 GB with a q8 KV cache) | Modified MIT | 1T MoE | Moonshot AI | — | — |
| 40 | [Kimi K2.7 Code](https://llmcheck.net/models/kimi-k27-code-on-m5-ultra/) | 67 | 27 | estimated | M5 Ultra | IQ3_S | 421 GB | 426 GB (estimated; 424 GB with a q8 KV cache) | Modified MIT | 1T MoE | Moonshot AI | — | — |
| 41 | [Ling-3.1-flash](https://llmcheck.net/models/ling-31-flash-on-m5-ultra/) | 67 | 27 | estimated | M5 Ultra | Q4_K_M | 360 GB | 361 GB (estimated; 361 GB with a q8 KV cache) | MIT | 560B MoE | Ant Group | — | — |
| 42 | [Muse Glimmer 30B](https://llmcheck.net/models/muse-glimmer-30b-on-m5-max/) | 67 | 27 | sourced | M5 Max | Q4_K_M | 18 GB | 19 GB (estimated; 19 GB with a q8 KV cache) | Apache 2.0 | 30B | Meta | TTFT 0.4 s (sourced) | — |
| 43 | [Qwen3.8-Flash-Next](https://llmcheck.net/models/qwen38-flash-next-on-m5-max/) | 67 | 47 | community | M5 Max | 4-bit | 120 GB | 122 GB (estimated; 121 GB with a q8 KV cache) | Qwen Custom | 125B MoE | Alibaba | prefill 1,309 tok/s @1K (community, M5 Max) | — |
| 44 | [GLM-5.3](https://llmcheck.net/models/glm-53-on-m5-ultra/) | 66 | 17 | estimated | M5 Ultra | Q4_K_M | 256 GB | 262 GB (estimated; 259 GB with a q8 KV cache) | GLM-5.3 | 753B MoE | Zhipu AI | — | — |
| 45 | [Gemma 4 12B](https://llmcheck.net/models/gemma-4-12b-on-m5-max/) | 66 | 64 | estimated | M5 Max | Q4_K_M | 9 GB | 11 GB (estimated; 10 GB with a q8 KV cache) | Apache 2.0 | 12B | Google | — | `gemma4:12b` |
| 46 | Kimi K2.5 | 66 | 28 | estimated | M5 Ultra | 3.5 bpw | 417 GB | 422 GB (estimated; 420 GB with a q8 KV cache) | Modified MIT | 1T MoE | Moonshot AI | — | — |
| 47 | [MiMo-V2.6-Distill-Qwen-9B](https://llmcheck.net/models/mimo-v26-distill-qwen-9b-on-m5-max/) | 66 | 79 | estimated | M5 Max | Q4_K_M | 8 GB | 11 GB (estimated; 10 GB with a q8 KV cache) | MIT | 9.41B | Xiaomi | — | `mimo-v2.6:9b` |
| 48 | [MiniMax M2.5](https://llmcheck.net/models/minimax-m25-on-m5-ultra/) | 66 | 57 | estimated | M5 Ultra | Q4_K_M | 130 GB | 146 GB (estimated; 138 GB with a q8 KV cache) | Modified MIT | 230B MoE | MiniMax | — | — |
| 49 | [Beam 1.0](https://llmcheck.net/models/beam-10-on-m5-ultra/) | 65 | 27 | estimated | M5 Ultra | Q4_K_M | 199 GB | 202 GB (estimated; 201 GB with a q8 KV cache) | Apache 2.0 | 501.3B MoE | Reflection AI | — | — |
| 50 | GLM-5.1 | 65 | 17 | estimated | M5 Ultra | Q4_K_M | 390 GB | 396 GB (estimated; 393 GB with a q8 KV cache) | MIT | 754B MoE | Zhipu AI | — | — |
| 51 | [Ling-3.0-tiny](https://llmcheck.net/models/ling-30-tiny-on-m5-max/) | 65 | 174 | community | M5 Max | oQ4e | 6 GB | 7 GB (estimated; 7 GB with a q8 KV cache) | MIT | 7.9B MoE | Ant Group | prefill 2,012 tok/s @4K (community, M5 Max); TTFT 2.035 s @4K (community) | — |
| 52 | [MiniMax-M3](https://llmcheck.net/models/minimax-m3-on-m5-ultra/) | 65 | 42 | estimated | M5 Ultra | Q4_K_M | 162 GB | 170 GB (estimated; 166 GB with a q8 KV cache) | MiniMax Community | 428B MoE | MiniMax | — | — |
| 53 | [Muse Spark Open](https://llmcheck.net/models/muse-spark-open-on-m5-ultra/) | 65 | 31 | estimated | M5 Ultra | oQ3e | 342 GB | — | Muse Model License | 816B MoE | Meta | — | — |
| 54 | [Nemotron 3.5 Bolt E4B](https://llmcheck.net/models/nemotron-35-bolt-e4b-on-m5-max/) | 65 | 161 | estimated | M5 Max | Q4_K_M | 6 GB | 8 GB (estimated; 7 GB with a q8 KV cache) | OpenMDW | 4B | NVIDIA | — | `nemotron-3.5-bolt:e4b` |
| 55 | [Qwen 3 30B-A3B](https://llmcheck.net/models/qwen-3-30b-a3b-on-m5-max/) | 65 | 152 | sourced | M5 Max | 4-bit | 18 GB | 25 GB (estimated; 22 GB with a q8 KV cache) | Apache 2.0 | 30B MoE | Alibaba | — | `qwen3:30b-a3b` |
| 56 | Qwen 3.5 397B-A17B | 65 | 62 | community | M5 Ultra | 4-bit | 210 GB | 212 GB (estimated; 211 GB with a q8 KV cache) | Apache 2.0 | 397B MoE | Alibaba | prefill 1,265 tok/s @8K (community, M5 Ultra) | — |
| 57 | [DeepSeek R1 8B](https://llmcheck.net/models/deepseek-r1-8b-on-m6/) | 64 | 91 | estimated | M5 Max | Q4_K_M | 6 GB | 16 GB (estimated; 11 GB with a q8 KV cache) | MIT | 8B | DeepSeek | — | `deepseek-r1:8b` |
| 58 | [Gemma 4 31B](https://llmcheck.net/models/gemma-4-31b-on-m5-max/) | 64 | 26 | community | M5 Max | 4-bit | 24 GB | 30 GB (estimated; 27 GB with a q8 KV cache) | Apache 2.0 | 31B | Google | prefill 710 tok/s @4K (community, M5 Max); TTFT 1.439 s @1K (community) | `gemma4:31b` |
| 59 | [Gemma 4 E4B](https://llmcheck.net/models/gemma-4-e4b-on-m5-max/) | 64 | 161 | estimated | M5 Max | Q4_K_M | 6 GB | 8 GB (estimated; 7 GB with a q8 KV cache) | Apache 2.0 | 4B | Google | — | `gemma4:e4b` |
| 60 | [Laguna S 2.1](https://llmcheck.net/models/laguna-s-21-on-m5-ultra/) | 64 | 38 | estimated | M5 Max | Q4_K_M | 96 GB | 100 GB (estimated; 98 GB with a q8 KV cache) | OpenMDW | 118B MoE | Poolside | — | — |
| 61 | [MiniMax-M3.1-Flash](https://llmcheck.net/models/minimax-m31-flash-on-m5-ultra/) | 64 | 35 | community | M5 Ultra | oQ3e | 189 GB | 197 GB (estimated; 193 GB with a q8 KV cache) | MiniMax Community | 428B MoE | MiniMax | prefill 1,110 tok/s @4K (community, M5 Ultra); TTFT 0.969 s @1K (community) | — |
| 62 | Phi-4 Mini | 64 | 168 | estimated | M5 Max | Q4_K_M | 3 GB | 12 GB (estimated; 8 GB with a q8 KV cache) | MIT | 3.8B | Microsoft | — | `phi4-mini:3.8b` |
| 63 | [Step-3.7-Flash](https://llmcheck.net/models/step-37-flash-on-m5-max/) | 64 | 48 | community | M5 Max | oQ2e | 65 GB | 69 GB (estimated; 67 GB with a q8 KV cache) | Apache 2.0 | 198B MoE | StepFun | prefill 594 tok/s @4K (community, M5 Max); TTFT 1.704 s @1K (community) | — |
| 64 | DeepSeek V4 Pro | 63 | — | fits no single Mac | — | — | 850 GB | 851 GB (estimated; 851 GB with a q8 KV cache) | MIT | 1.6T MoE | DeepSeek | — | — |
| 65 | [Hy4 preview](https://llmcheck.net/models/hy4-preview-on-m5-ultra/) | 63 | 14 | estimated | M5 Ultra | Q4_K_M | 292 GB | 298 GB (estimated; 295 GB with a q8 KV cache) | Apache 2.0 | 770B MoE | Tencent | — | — |
| 66 | [LFM3-1.7B](https://llmcheck.net/models/lfm3-17b-on-m5-max/) | 63 | 285 | estimated | M5 Max | Q4_K_M | 3 GB | 4 GB (estimated; 4 GB with a q8 KV cache) | LFM Open | 1.73B | Liquid AI | — | `lfm3:1.7b` |
| 67 | [Mellum2.1-12B-A2.5B-Thinking](https://llmcheck.net/models/mellum21-12b-a25b-thinking-on-m5-max/) | 63 | 154 | estimated | M5 Max | Q4_K_M | 10 GB | 11 GB (estimated; 11 GB with a q8 KV cache) | Apache 2.0 | 12.15B MoE | JetBrains | — | — |
| 68 | MiMo-V2.6-Pro | 63 | — | fits no single Mac | — | — | 581 GB | 585 GB (estimated; 583 GB with a q8 KV cache) | MIT | 1.02T MoE | Xiaomi | — | — |
| 69 | [Qwen3-235B-A22B](https://llmcheck.net/models/qwen3-235b-a22b-on-m5-ultra/) | 63 | 26 | estimated | M5 Max | 4.5 bpw | 128 GB | 140 GB (estimated; 134 GB with a q8 KV cache) | Apache 2.0 | 235B MoE | Alibaba | — | `qwen3:235b-a22b` |
| 70 | Apertus 1.5 8B | 62 | 91 | estimated | M5 Max | Q4_K_M | 5 GB | — | Apache 2.0 | 8B | Swiss AI | — | — |
| 71 | DeepSeek V3.2-Speciale | 62 | 18 | estimated | M5 Ultra | Q4_K_M | 360 GB | 365 GB (estimated; 363 GB with a q8 KV cache) | MIT | 685B MoE | DeepSeek | — | — |
| 72 | [Nemotron 3.5 Super VL 120B-A12B](https://llmcheck.net/models/nemotron-35-super-vl-120b-a12b-on-m5-max/) | 62 | 36 | community | M5 Max | Q4_0 | 72 GB | 73 GB (estimated; 73 GB with a q8 KV cache) | OpenMDW | 120B MoE | NVIDIA | prefill 885 tok/s @4K (community, M5 Max) | `nemotron-3.5-super:120b` |
| 73 | [dots3-note preview](https://llmcheck.net/models/dots3-note-preview-on-m5-max/) | 62 | 28 | estimated | M5 Max | 3.5 bpw | 120 GB | 121 GB (estimated; 121 GB with a q8 KV cache) | Apache 2.0 | 280B MoE | Xiaohongshu (RedNote) | — | — |
| 74 | [Gemma 4 E2B](https://llmcheck.net/models/gemma-4-e2b-on-m5-max/) | 61 | 261 | estimated | M5 Max | Q4_K_M | 2 GB | 3 GB (estimated; 3 GB with a q8 KV cache) | Apache 2.0 | 2B | Google | — | `gemma4:e2b` |
| 75 | Mistral 7B | 61 | 102 | estimated | M5 Max | Q4_K_M | 5 GB | 32K window (below 64K) | Apache 2.0 | 7B | Mistral | — | `mistral:7b` |
| 76 | [Qwen 3 8B](https://llmcheck.net/models/qwen-3-8b-on-m5-max/) | 61 | 91 | estimated | M5 Max | Q4_K_M | 6 GB | 16 GB (estimated; 11 GB with a q8 KV cache) | Apache 2.0 | 8B | Alibaba | — | `qwen3:8b` |
| 77 | [Qwen 3.5 9B](https://llmcheck.net/models/qwen-35-9b-on-m5-max/) | 61 | 81 | estimated | M5 Max | Q4_K_M | 7 GB | 10 GB (estimated; 9 GB with a q8 KV cache) | Apache 2.0 | 9B | Alibaba | — | `qwen3.5:9b` |
| 78 | [Granite 4.2 8B](https://llmcheck.net/models/granite-42-8b-on-m5-max/) | 60 | 84 | estimated | M5 Max | Q4_K_M | 8 GB | 19 GB (estimated; 14 GB with a q8 KV cache) | Apache 2.0 | 8.79B | IBM | — | `granite4.2:8b` |
| 79 | Hunyuan Hy3 | 60 | 30 | estimated | M5 Ultra | Q4_K_M | 160 GB | 181 GB (estimated; 171 GB with a q8 KV cache) | Apache 2.0 | 295B MoE | Tencent | — | — |
| 80 | [LFM2.5-2.6B](https://llmcheck.net/models/lfm25-26b-on-m5-max/) | 60 | 220 | sourced | M5 Max | Q4_K_M | 2 GB | 4 GB (estimated; 3 GB with a q8 KV cache) | LFM Open | 2.6B | Liquid AI | TTFT 0.1 s (sourced) | — |
| 81 | Ministral 8B | 60 | 91 | estimated | M5 Max | Q4_K_M | 6 GB | 12 GB (estimated; 9 GB with a q8 KV cache) | Apache 2.0 | 8B | Mistral | — | — |
| 82 | [Qwen 3.5 122B-A10B](https://llmcheck.net/models/qwen-35-122b-a10b-on-m5-max/) | 60 | 43 | community | M5 Max | Q4_K_M | 65 GB | 67 GB (estimated; 66 GB with a q8 KV cache) | Apache 2.0 | 122B MoE | Alibaba | — | `qwen3.5:122b-a10b` |
| 83 | [Qwen 3.5 4B](https://llmcheck.net/models/qwen-35-4b-on-m5-max/) | 60 | 161 | estimated | M5 Max | Q4_K_M | 3 GB | 6 GB (estimated; 5 GB with a q8 KV cache) | Apache 2.0 | 4B | Alibaba | — | `qwen3.5:4b` |
| 84 | DeepSeek V3.2 | 59 | 18 | estimated | M5 Ultra | Q4_K_M | 360 GB | 365 GB (estimated; 363 GB with a q8 KV cache) | MIT | 685B MoE | DeepSeek | — | — |
| 85 | [Mistral Small 4](https://llmcheck.net/models/mistral-small-4-on-m5-ultra/) | 59 | 46 | estimated | M5 Max | Q4_K_M | 64 GB | 66 GB (estimated; 65 GB with a q8 KV cache) | Apache 2.0 | 119B MoE | Mistral | — | — |
| 86 | [Qwen 3 4B](https://llmcheck.net/models/qwen-3-4b-on-m6/) | 59 | 161 | estimated | M5 Max | Q4_K_M | 3 GB | 13 GB (estimated; 8 GB with a q8 KV cache) | Apache 2.0 | 4B | Alibaba | — | `qwen3:4b` |
| 87 | [Ripple-65B Preview](https://llmcheck.net/models/ripple-65b-preview-on-m5-max/) | 59 | 13 | estimated | M5 Max | Q4_K_M | 33 GB | 38 GB (estimated; 36 GB with a q8 KV cache) | Apache 2.0 | 64.8B | Ricursive Intelligence | — | `ripple-preview:65b` |
| 88 | Inkling | 58 | — | fits no single Mac | — | — | 550 GB | 553 GB (estimated; 552 GB with a q8 KV cache) | Apache 2.0 | 975B MoE | Thinking Machines | — | — |
| 89 | Kimi K3 | 58 | — | fits no single Mac | — | — | 1560 GB | 1562 GB (estimated; 1561 GB with a q8 KV cache) | Moonshot Custom | 2.8T MoE | Moonshot AI | — | — |
| 90 | Kimi K3.1 Pro | 58 | — | fits no single Mac | — | — | 1563 GB | 1565 GB (estimated; 1564 GB with a q8 KV cache) | Moonshot Custom | 2.8T MoE | Moonshot AI | — | — |
| 91 | [MiniMax-M2.7](https://llmcheck.net/models/minimax-m27-on-m5-max/) | 58 | 45 | estimated | M5 Max | IQ3_XXS | 82 GB | 98 GB (estimated; 90 GB with a q8 KV cache) | MiniMax Non-Commercial | 228.7B MoE | MiniMax | — | — |
| 92 | [Mistral Large 4](https://llmcheck.net/models/mistral-large-4-on-m5-ultra/) | 58 | 21 | estimated | M5 Ultra | 3.0 bpw | 383 GB | 388 GB (estimated; 386 GB with a q8 KV cache) | Mistral Ecosystem Open-Weight | 1.05T MoE | Mistral | — | — |
| 93 | Qwen3.8-2.4T-A95B | 58 | — | fits no single Mac | — | — | 1310 GB | 1316 GB (estimated; 1313 GB with a q8 KV cache) | Qwen Custom | 2.4T MoE | Alibaba | — | — |
| 94 | SmolLM3 3B | 58 | 199 | estimated | M5 Max | Q4_K_M | 2 GB | 7 GB (estimated; 5 GB with a q8 KV cache) | Apache 2.0 | 3B | HuggingFace | — | — |
| 95 | [Ternary Bonsai 2 9B](https://llmcheck.net/models/ternary-bonsai-2-9b-on-m5-pro/) | 58 | 77 | sourced | M5 Pro | PQ2_0 | 5 GB | 8 GB (estimated; 7 GB with a q8 KV cache) | Apache 2.0 | 9.41B | Prism ML | — | `bonsai-2:9b` |
| 96 | [Llama 3.1 8B](https://llmcheck.net/models/llama-31-8b-on-m5-max/) | 57 | 102 | sourced | M5 Max | 4-bit | 5 GB | — | Meta Custom | 8B | Meta | — | `llama3.1:8b` |
| 97 | Step 5 Preview | 57 | — | no Mac runtime yet | — | — | 261 GB | 266 GB (estimated; 264 GB with a q8 KV cache) | StepFun Model License | 600B MoE | StepFun | — | — |
| 98 | [Gemma 3 4B](https://llmcheck.net/models/gemma-3-4b-on-m3-pro/) | 56 | 161 | estimated | M5 Max | Q4_K_M | 3 GB | — | Gemma | 4B | Google | — | `gemma3:4b` |
| 99 | [Granite 4.2 30B](https://llmcheck.net/models/granite-42-30b-on-m5-max/) | 56 | 28 | estimated | M5 Max | Q4_K_M | 20 GB | 37 GB (estimated; 29 GB with a q8 KV cache) | Apache 2.0 | 29.28B | IBM | — | `granite4.2:30b` |
| 100 | [Nemotron 3 Ultra 550B-A55B](https://llmcheck.net/models/nemotron-3-ultra-550b-a55b-on-m5-ultra/) | 56 | 12 | estimated | M5 Ultra | Q4_K_M | 226 GB | 227 GB (estimated; 227 GB with a q8 KV cache) | OpenMDW | 550B MoE | NVIDIA | — | — |
| 101 | [GLM-4.5-Air](https://llmcheck.net/models/glm-45-air-on-m5-ultra/) | 55 | 27 | estimated | M5 Max | Q4_K_M | 56 GB | 68 GB (estimated; 62 GB with a q8 KV cache) | MIT | 106B MoE | Zhipu AI | — | — |
| 102 | MiMo-V2-Flash | 55 | 41 | estimated | M5 Ultra | Q4_K_M | 165 GB | 167 GB (estimated; 166 GB with a q8 KV cache) | MIT | 309B MoE | Xiaomi | — | — |
| 103 | [Phi-4 14B](https://llmcheck.net/models/phi-4-14b-on-m6/) | 55 | 55 | estimated | M5 Max | Q4_K_M | 10 GB | 23 GB (estimated; 17 GB with a q8 KV cache) | MIT | 14B | Microsoft | — | `phi4:14b` |
| 104 | DeepSeek R1 | 54 | 18 | estimated | M5 Ultra | Q4_K_M | 350 GB | 355 GB (estimated; 353 GB with a q8 KV cache) | MIT | 671B MoE | DeepSeek | — | `deepseek-r1:671b` |
| 105 | DeepSeek V3 | 54 | 18 | estimated | M5 Ultra | Q4_K_M | 360 GB | 365 GB (estimated; 363 GB with a q8 KV cache) | MIT | 685B MoE | DeepSeek | — | `deepseek-v3:671b` |
| 106 | [Qwen 3 14B](https://llmcheck.net/models/qwen-3-14b-on-m4-pro/) | 54 | 55 | estimated | M5 Max | Q4_K_M | 10 GB | 21 GB (estimated; 16 GB with a q8 KV cache) | Apache 2.0 | 14B | Alibaba | — | `qwen3:14b` |
| 107 | Llama 4 Maverick | 53 | 36 | estimated | M5 Ultra | Q4_K_M | 210 GB | — | Meta Custom | 400B MoE | Meta | — | `llama4:maverick` |
| 108 | [Mistral Medium 3.5](https://llmcheck.net/models/mistral-medium-35-on-m5-max/) | 53 | 7 | estimated | M5 Max | Q4_K_M | 77 GB | 100 GB (estimated; 89 GB with a q8 KV cache) | Mistral Modified MIT | 128B | Mistral | — | `mistral-medium-3.5:128b` |
| 109 | Mistral Small 3.2 24B | 53 | 33 | estimated | M5 Max | Q4_K_M | 14 GB | 25 GB (estimated; 20 GB with a q8 KV cache) | Apache 2.0 | 24B | Mistral | — | `mistral-small3.2:24b` |
| 110 | Devstral Small 24B | 52 | 33 | estimated | M5 Max | Q4_K_M | 14 GB | 25 GB (estimated; 20 GB with a q8 KV cache) | Apache 2.0 | 24B | Mistral | — | — |
| 111 | GLM-4.7 | 52 | 21 | estimated | M5 Ultra | Q4_K_M | 180 GB | 204 GB (estimated; 192 GB with a q8 KV cache) | MIT | 355B MoE | Zhipu AI | — | — |
| 112 | Ministral 3 14B | 52 | 55 | estimated | M5 Max | Q4_K_M | 10 GB | 21 GB (estimated; 16 GB with a q8 KV cache) | Apache 2.0 | 14B | Mistral | — | `ministral-3:14b` |
| 113 | Qwen 2.5 14B | 52 | 55 | estimated | M5 Max | Q4_K_M | 10 GB | 23 GB (estimated; 17 GB with a q8 KV cache) | Apache 2.0 | 14B | Alibaba | — | `qwen2.5:14b` |
| 114 | [Gemma 3 12B](https://llmcheck.net/models/gemma-3-12b-on-m5-max/) | 51 | 64 | estimated | M5 Max | Q4_K_M | 9 GB | — | Gemma | 12B | Google | — | `gemma3:12b` |
| 115 | Mistral Large 3 | 51 | 16 | estimated | M5 Ultra | Q4_K_M | 355 GB | 360 GB (estimated; 358 GB with a q8 KV cache) | Apache 2.0 | 675B MoE | Mistral | — | — |
| 116 | Solar Open 2 250B | 51 | — | no Mac runtime yet | — | — | 160 GB | 164 GB (estimated; 162 GB with a q8 KV cache) | Solar License | 250B MoE | Upstage | — | — |
| 117 | Step-3.5-Flash | 51 | 29 | estimated | M5 Max | Q4_K_M | 120 GB | 124 GB (estimated; 122 GB with a q8 KV cache) | Apache 2.0 | 196B MoE | StepFun | — | — |
| 118 | DeepSeek R1 32B | 50 | 25 | estimated | M5 Max | Q4_K_M | 22 GB | 39 GB (estimated; 31 GB with a q8 KV cache) | MIT | 32B | DeepSeek | — | `deepseek-r1:32b` |
| 119 | QwQ 32B | 50 | 25 | estimated | M5 Max | Q4_K_M | 22 GB | 32K window (below 64K) | Apache 2.0 | 32B | Alibaba | — | `qwq:32b` |
| 120 | [DeepSeek R1 70B](https://llmcheck.net/models/deepseek-r1-70b-on-m5-max/) | 49 | 12 | estimated | M5 Max | Q4_K_M | 44 GB | 65 GB (estimated; 55 GB with a q8 KV cache) | MIT | 70B | DeepSeek | — | `deepseek-r1:70b` |
| 121 | [Qwen 3 32B](https://llmcheck.net/models/qwen-3-32b-on-m5-max/) | 49 | 26 | sourced | M5 Max | 4-bit | 22 GB | 39 GB (estimated; 31 GB with a q8 KV cache) | Apache 2.0 | 32B | Alibaba | — | `qwen3:32b` |
| 122 | [Qwen 3.5 27B](https://llmcheck.net/models/qwen-35-27b-on-m5-ultra/) | 47 | 30 | estimated | M5 Max | Q4_K_M | 18 GB | 23 GB (estimated; 21 GB with a q8 KV cache) | Apache 2.0 | 27B | Alibaba | — | `qwen3.5:27b` |
| 123 | Apertus 1.5 70B | 46 | 12 | estimated | M5 Max | Q4_K_M | 44 GB | — | Apache 2.0 | 70B | Swiss AI | — | — |
| 124 | [Gemma 3 27B](https://llmcheck.net/models/gemma-3-27b-on-m4-max/) | 46 | 30 | estimated | M5 Max | Q4_K_M | 18 GB | — | Gemma | 27B | Google | — | `gemma3:27b` |
| 125 | Llama 4 Scout | 45 | 19 | estimated | M5 Max | Q4_K_M | 58 GB | — | Meta Custom | 109B MoE | Meta | — | `llama4:scout` |
| 126 | [Qwen 2.5 72B](https://llmcheck.net/models/qwen-25-72b-on-m5-ultra/) | 45 | 12 | estimated | M5 Max | Q4_K_M | 46 GB | 67 GB (estimated; 57 GB with a q8 KV cache) | Apache 2.0 | 72B | Alibaba | — | `qwen2.5:72b` |
| 127 | [Llama 3.1 405B](https://llmcheck.net/models/llama-31-405b-on-m5-ultra/) | 44 | 4 | estimated | M5 Ultra | Q4_K_M | 240 GB | — | Meta Custom | 405B | Meta | — | `llama3.1:405b` |
| 128 | Hermes 4 70B | 42 | 12 | estimated | M5 Max | Q4_K_M | 38 GB | 59 GB (estimated; 49 GB with a q8 KV cache) | Meta Custom | 70B | Nous Research | — | — |
| 129 | [Llama 3.3 70B](https://llmcheck.net/models/llama-33-70b-on-m5-max/) | 41 | 13 | sourced | M5 Max | 4-bit | 44 GB | — | Meta Custom | 70B | Meta | prefill 306 tok/s @0.5K (sourced, M5 Max) | `llama3.3:70b` |
| 130 | Mixtral 8x22B | 41 | 9 | estimated | M5 Max | Q4_K_M | 88 GB | 103 GB (estimated; 96 GB with a q8 KV cache) | Apache 2.0 | 141B MoE | Mistral | — | `mixtral:8x22b` |

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).

Measured figure: 35 of 130 rows (sourced or community); verified Ollama tag: 61 of 130. Method: https://llmcheck.net/methodology#reference-row · Context and memory: https://llmcheck.net/methodology#context · Why speculative-decoding (MTP) speeds never set a rank: https://llmcheck.net/methodology#speculative

## Questions about the ranking

### What is the best local LLM for a Mac right now?

According to the LLMCheck index, as of 18 November 2026, MiMo-V2.6-Flash and Qwen 3.6-35B-A3B tie for #1 at LLMCheck Score 81; MiMo-V2.6-Flash is listed first on the alphabetical tie-break. MiMo-V2.6-Flash runs at 77 tok/s community-measured on an M5 Ultra but needs a Mac Studio with at least 192 GB of unified memory. Qwen 3.6-35B-A3B needs only 24 GB and is the best pick on a Mac with 32 GB or less, at 59 tok/s on a Mac mini M6 32 GB (community-measured on an M6). The full ranking of all 130 models is at llmcheck.net/leaderboard.

### What is the best local LLM for each amount of RAM?

According to the LLMCheck index, as of 18 November 2026, the best pick for each RAM tier is the best-ranked model that fits the tier, reaches 15 tok/s on the tier's representative Mac and has a standard Mac runtime (the /best-llm/by-ram/ pick), ranking capability against speed: 8 GB Mac: MiniCPM5-2B (Score 69; ~39 tok/s estimated on a MacBook Air M1 8 GB; needs 3 GB); 16 GB Mac: Gemma 4 12B (Score 66; ~17 tok/s estimated on a Mac mini M6 16 GB; needs 9 GB); 24 GB Mac: Recursive Flint (Score 80; ~71 tok/s estimated on a Mac mini M6 24 GB; needs 20 GB); 32 GB Mac: Qwen 3.6-35B-A3B (Score 81; 59 tok/s on a Mac mini M6 32 GB (community-measured on an M6); needs 24 GB); 64 GB Mac: LFM2.5-24B-A2B (Score 76; 114 tok/s on a Mac mini M5 Pro 64 GB (community-measured on an M5 Pro); needs 17 GB); 128 GB Mac: Qwen 3.6-35B-A3B (Score 81; 130 tok/s on a Mac Studio M5 Max 128 GB (community-measured on an M5 Max); needs 24 GB); 192–256 GB Mac: MiMo-V2.6-Flash (Score 81; 77 tok/s on a Mac Studio M5 Ultra 256 GB (community-measured on an M5 Ultra); needs 170 GB); 512 GB Mac: MiMo-V2.6-Flash (Score 81; 77 tok/s on a Mac Studio M5 Ultra 512 GB (community-measured on an M5 Ultra); needs 170 GB). Each tier, with every Mac in it: llmcheck.net/best-llm/by-ram/.

### What is the best dense local LLM?

The strongest dense (non-mixture-of-experts) model in the index is Qwen3.8-27B (capability 46/50, 19 GB minimum, 31 tok/s community-measured on an M5 Max at 4-bit). Dense models read every weight for every token, so they decode slower than mixture-of-experts models of similar size.

### What is the fastest local LLM on a Mac?

The fastest measured figure in the index is Maple Preview 20B-A1B at 387 tok/s on an M5 Pro (published, Ternary); its ternary weights decode through compute-bound kernels, so among bandwidth-bound models the fastest measured reference figure is LFM2.5-2.6B at 220 tok/s published on an M5 Max (Q4_K_M).

### How is the LLMCheck Score calculated?

Capability (0-50, normalized from published benchmarks) + speed (0-25: the displayed tok/s x 0.25, capped) + RAM accessibility (0-15: 15 at 8 GB or less, down to 3 above 64 GB) + licence openness (0-10: MIT 10 down to non-commercial 2). The speed figure is the lower-median measured plain-decode run on an M5 Max (or an M5 Ultra for models above 128 GB) where one exists, otherwise the published bandwidth estimate. Method: llmcheck.net/methodology#reference-row.

### Why did models move in the 19 November 2026 update?

The 19 November 2026 update (data of 18 November) changed no ranking rule. It added a RAM-at-64K figure for each model whose KV-cache layout has been read from its config.json (minimum RAM covers about the first 4K tokens of context; a longer context adds its cache), decode-at-depth and prefill figures where a run measured them, and speculative-decoding (MTP) figures kept beside their paired plain run. None of these enters the Score. As of 18 November 2026, 320 measured rows cover 61 of the 130 models (all listed at llmcheck.net/benchmarks#measured); for 35 of the 130 models the leaderboard figure itself is measured, and the rest show an estimate on their reference chip. The largest gaps between a measurement and the formula's own estimate at the same quant: Nemotron 3.5 Bolt 74B-A9B 64 tok/s community-measured vs 30 estimated on an M5 Max (5-bit); MiMo-V2.6-Flash 77 tok/s community-measured vs 42 estimated on an M5 Ultra (MXFP4); Qwen 3.5 397B-A17B 62 tok/s community-measured vs 36 estimated on an M5 Ultra (4-bit). GLM-4.7-Flash's M5 Max figure (~107 tok/s, estimated) is scaled from its own community-measured M6 run of 46.6 tok/s. Every change is logged at llmcheck.net/changelog.

### How much memory does a 64K context need?

According to the LLMCheck index, a model's RAM at 64K is its minimum RAM, which covers about the first 4K tokens, plus the 16-bit KV cache the rest of a 65,536-token context adds, computed from the attention layout in its config.json: an estimate, not a measurement. Qwen 3.6-35B-A3B needs 24 GB at a short context and about 26 GB at 64K (25 GB with an 8-bit KV cache). Sliding-window and linear-attention layers add less. Of the 117 with a recorded layout, 51 are agent-ready at 64K on a 32 GB Mac. Method: llmcheck.net/methodology#context.


## Cite this page

LLMCheck. (2026). The local LLM leaderboard. llmcheck.net. https://llmcheck.net/leaderboard (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
