---
title: "Models vs memory"
schema_version: 1.0
generated: 2026-10-15
json: https://insidethemachine.ai/models.json
count: 124
count_note: "text models only; speech models in https://insidethemachine.ai/speech.json (40), image/video models in https://insidethemachine.ai/images.json (30)"
---

# Models vs memory

## Methodology

- **memory_formula**: min_memory_gb = weights_gb_4bit × 1.15 + 2. weights_gb_4bit = Q4_K_M file (Ollama, else HF GGUF) → MLX 4-bit (HF) → Ollama MLX/NVFP4 tag → Ollama default tag → params_total_b × 0.6 (flagged weights_estimated). +15% ≈ KV cache at ~8–16K context; +2GB runtime/process overhead. Long contexts (128K+) need much more.
- **tier_rule**: A model fits a tier when min_memory_gb ≤ 0.75 × tier_gb. 0.75 approximates the default share of unified memory macOS lets the GPU wire (can be raised with `sudo sysctl iogpu.wired_limit_mb=<MB>`), and leaves room for OS/apps on PCs. For discrete GPUs, treat tier_gb as VRAM (CPU offload is possible but slow).
- **file_sizes**: Copied as listed on Ollama tag pages (GB, rounded by Ollama) or summed from Hugging Face repo file sizes (bytes/1e9). Retrieved 2026-10-15.
- **benchmarks**: independent = Artificial Analysis (current index version as displayed 2026-10-15; not comparable to older AA index numbers); vendor-reported = parsed from the model card table on Hugging Face (column matched to this model; verify before quoting); crowd-preference = LMArena Text leaderboard Elo (2026-10-15).
- **gap**: Derived only: AA points behind the top closed model, and Elo behind LMArena #1. Not the site's Gap Score (that needs Mac Studio tests).

## Picks by RAM tier

| tier GB | usable GB | fits | top picks | best coding | best agentic | embedding |
|---|---|---|---|---|---|---|
| 8 | 6 | 19 | qwen3.5-4b, minicpm5-2b, granite-4.2-3b | - | - | embeddinggemma-2 |
| 16 | 12 | 38 | k2-horizon-7b, gemma-4-12b, qwen3.5-9b | ornith-1.5-9b | mellum2.1-12b-a2.5b | qwen3-embedding-8b |
| 24 | 18 | 45 | k2-horizon-7b, gemma-4-12b, qwen3.5-9b | nouscoder-14b | mellum2.1-12b-a2.5b | qwen3-embedding-8b |
| 32 | 24 | 62 | qwen3.8-27b, qwen3.5-27b, qwen3.6-27b | qwen3.8-27b | qwen3.8-27b | qwen3-embedding-8b |
| 48 | 36 | 76 | qwen3.8-27b, k2-horizon-mova-36b-a4b, qwen3.5-27b | qwen3.8-27b | qwen3.8-27b | qwen3-embedding-8b |
| 64 | 48 | 77 | qwen3.8-27b, k2-horizon-mova-36b-a4b, qwen3.5-27b | qwen3.8-27b | qwen3.8-27b | qwen3-embedding-8b |
| 96 | 72 | 83 | qwen3.8-27b, k2-horizon-mova-36b-a4b, qwen3.5-27b | qwen3.8-27b | qwen3.8-27b | qwen3-embedding-8b |
| 128 | 96 | 90 | qwen3.8-27b, k2-horizon-mova-36b-a4b, qwen3.5-27b | qwen3.8-27b | qwen3.8-27b | qwen3-embedding-8b |
| 192 | 144 | 94 | qwen3.8-flash-next, qwen3.8-27b, nemotron-3.5-super-120b | qwen3.8-flash-next | qwen3.8-flash-next | qwen3-embedding-8b |
| 256 | 192 | 95 | qwen3.8-flash-next, qwen3.8-27b, nemotron-3.5-super-120b | qwen3.8-flash-next | qwen3.8-flash-next | qwen3-embedding-8b |
| 512 | 384 | 102 | glm-5.3-flash, qwen3.8-flash-next, mimo-v2.6-flash | glm-5.3-flash | glm-5.3-flash | qwen3-embedding-8b |

## Models

| id | params B (active) | min memory GB | AA index | LMArena Elo | license | ollama | mlx |
|---|---|---|---|---|---|---|---|
| qwen3.8-27b | 27.78 | 22.7 | 33.7 | 1438 | apache-2.0 | `ollama run qwen3.8:27b` | `mlx_lm.generate --model mlx-community/Qwen3.8-27B-4bit --prompt 'hello'` |
| qwen3.8-flash-next | 180 (6) | 140 | 39.8 | - | qwen-community-1.0 | `ollama run qwen3.8-flash-next:125b-a6b` | `mlx_lm.generate --model mlx-community/Qwen3.8-Flash-Next-4bit --prompt 'hello'` |
| qwen3.8-2.4t-a95b | 2446.18 (95) | 1689.9 (est.) | 39.9 | - | qwen3.8-max | - | - |
| qwen3.6-27b | 27.78 | 21.5 | 21.4 | - | apache-2.0 | `ollama run qwen3.6:27b` | `mlx_lm.generate --model mlx-community/Qwen3.6-27B-4bit --prompt 'hello'` |
| qwen3.6-35b-a3b | 35.95 (3) | 29.6 | 18.2 | - | apache-2.0 | `ollama run qwen3.6:35b-a3b` | `mlx_lm.generate --model mlx-community/Qwen3.6-35B-A3B-4bit --prompt 'hello'` |
| qwen3.5-0.8b | 0.87 | 2.6 | 6.1 | - | apache-2.0 | `ollama run qwen3.5:0.8b` | `mlx_lm.generate --model mlx-community/Qwen3.5-0.8B-4bit --prompt 'hello'` |
| qwen3.5-2b | 2.27 | 4.2 | 6.9 | - | apache-2.0 | `ollama run qwen3.5:2b` | `mlx_lm.generate --model mlx-community/Qwen3.5-2B-4bit --prompt 'hello'` |
| qwen3.5-4b | 4.66 | 5.8 | 13.1 | - | apache-2.0 | `ollama run qwen3.5:4b` | `mlx_lm.generate --model mlx-community/Qwen3.5-4B-4bit --prompt 'hello'` |
| qwen3.5-9b | 9.65 | 9.6 | 13.3 | - | apache-2.0 | `ollama run qwen3.5:9b` | `mlx_lm.generate --model mlx-community/Qwen3.5-9B-4bit --prompt 'hello'` |
| qwen3.5-27b | 27.78 | 21.5 | 22.9 | 1409 | apache-2.0 | `ollama run qwen3.5:27b` | `mlx_lm.generate --model mlx-community/Qwen3.5-27B-4bit --prompt 'hello'` |
| qwen3.5-35b-a3b | 35.95 (3) | 27.3 | 19.3 | 1394 | apache-2.0 | `ollama run qwen3.5:35b-a3b` | `mlx_lm.generate --model mlx-community/Qwen3.5-35B-A3B-4bit --prompt 'hello'` |
| qwen3.5-122b-a10b | 125.09 (10) | 95.1 | 17.7 | 1416 | apache-2.0 | `ollama run qwen3.5:122b-a10b` | `mlx_lm.generate --model mlx-community/Qwen3.5-122B-A10B-4bit --prompt 'hello'` |
| qwen3.5-397b-a17b | 403.4 (17) | 282.7 | 21.4 | 1442 | apache-2.0 | - | `mlx_lm.generate --model mlx-community/Qwen3.5-397B-A17B-4bit --prompt 'hello'` |
| qwen3-coder-30b-a3b | 30.53 (3) | 23.8 | 9.6 | - | apache-2.0 | `ollama run qwen3-coder:30b-a3b` | `mlx_lm.generate --model mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit --prompt 'hello'` |
| qwen3-coder-next | 79.67 (3) | 61.8 | 9.2 | - | apache-2.0 | `ollama run qwen3-coder-next` | `mlx_lm.generate --model mlx-community/Qwen3-Coder-Next-4bit --prompt 'hello'` |
| qwen3-coder-480b-a35b | 480.15 (35) | 335.5 | 11.9 | 1388 | apache-2.0 | `ollama run qwen3-coder:480b-a35b` | `mlx_lm.generate --model mlx-community/Qwen3-Coder-480B-A35B-Instruct-4bit --prompt 'hello'` |
| qwen3-next-80b-a3b | 81.32 (3) | 59.5 | 11.2 | 1399 | apache-2.0 | `ollama run qwen3-next:80b-a3b-instruct` | `mlx_lm.generate --model mlx-community/Qwen3-Next-80B-A3B-Instruct-4bit --prompt 'hello'` |
| qwen3-30b-a3b-2507 | 30.53 (3) | 23.8 | 9.8 | 1382 | apache-2.0 | `ollama run qwen3:30b-a3b-instruct-2507` | `mlx_lm.generate --model mlx-community/Qwen3-30B-A3B-Instruct-2507-4bit --prompt 'hello'` |
| qwen3-8b | 8.19 | 8 | 6 | - | apache-2.0 | `ollama run qwen3:8b` | `mlx_lm.generate --model mlx-community/Qwen3-8B-4bit --prompt 'hello'` |
| qwen3-vl-8b | 8.77 | 9 | 8.2 | - | apache-2.0 | `ollama run qwen3-vl:8b-instruct` | `mlx_lm.generate --model mlx-community/Qwen3-VL-8B-Instruct-4bit --prompt 'hello'` |
| qwen3-vl-30b-a3b | 31.07 (3) | 25 | 9.5 | - | apache-2.0 | `ollama run qwen3-vl:30b-a3b-instruct` | `mlx_lm.generate --model mlx-community/Qwen3-VL-30B-A3B-Instruct-4bit --prompt 'hello'` |
| qwen3-embedding-0.6b | 0.6 | 2.7 | - | - | apache-2.0 | `ollama run qwen3-embedding:0.6b` | - |
| qwen3-embedding-8b | 7.57 | 7.4 | - | - | apache-2.0 | `ollama run qwen3-embedding:8b` | - |
| gemma-4-e2b | 5.12 | 7.3 | 7.8 | - | apache-2.0 | `ollama run gemma4:e2b` | `mlx_lm.generate --model mlx-community/gemma-4-e2b-it-4bit --prompt 'hello'` |
| gemma-4-e4b | 8 | 9.6 | 8.9 | - | apache-2.0 | `ollama run gemma4:e4b` | `mlx_lm.generate --model mlx-community/gemma-4-e4b-it-4bit --prompt 'hello'` |
| gemma-4-12b | 11.96 | 11.2 | 14.2 | - | apache-2.0 | `ollama run gemma4:12b` | `mlx_lm.generate --model mlx-community/gemma-4-12B-it-4bit --prompt 'hello'` |
| gemma-4-26b-a4b | 25.81 (4) | 22.7 | 16.7 | 1438 | apache-2.0 | `ollama run gemma4:26b` | `mlx_lm.generate --model mlx-community/gemma-4-26b-a4b-it-4bit --prompt 'hello'` |
| gemma-4-31b | 31.27 | 25 | 14.7 | 1452 | apache-2.0 | `ollama run gemma4:31b` | `mlx_lm.generate --model mlx-community/gemma-4-31b-it-4bit --prompt 'hello'` |
| gemma-3-27b | 27.43 | 21.5 | 4.9 | 1365 | gemma | `ollama run gemma3:27b` | `mlx_lm.generate --model mlx-community/gemma-3-27b-it-4bit --prompt 'hello'` |
| gemma-3-4b | 4.3 | 5.9 | 4.8 | 1303 | gemma | `ollama run gemma3:4b` | `mlx_lm.generate --model mlx-community/gemma-3-4b-it-4bit --prompt 'hello'` |
| gemma-3n-e4b | 7.85 | 13.3 | 4.8 | 1318 | gemma | `ollama run gemma3n:e4b` | `mlx_lm.generate --model mlx-community/gemma-3n-E4B-it-4bit --prompt 'hello'` |
| functiongemma-270m | 0.27 | 2.2 | - | - | gemma | `ollama run functiongemma:270m` | `mlx_lm.generate --model mlx-community/functiongemma-270m-it-4bit --prompt 'hello'` |
| embeddinggemma-300m | 0.3 | 2.2 | - | - | gemma | `ollama run embeddinggemma:300m` | `mlx_lm.generate --model mlx-community/embeddinggemma-300m-4bit --prompt 'hello'` |
| embeddinggemma-2 | 0.74 | 3.5 | - | - | apache-2.0 | `ollama run embeddinggemma-2` | - |
| muse-glimmer-30b | 29.78 | 22.7 | 17.5 | 1425 | apache-2.0 | `ollama run muse-glimmer:30b` | `mlx_lm.generate --model mlx-community/Muse-Glimmer-30B-4bit --prompt 'hello'` |
| muse-spark-open-816b-a42b | 822.04 (42) | 533 | 38.9 | - | muse-model-license | - | `mlx_lm.generate --model mlx-community/Muse-Spark-Open-816B-A42B-4bit --prompt 'hello'` |
| llama-4-scout | 108.64 (17) | 79 | 6.5 | 1322 | llama4 | `ollama run llama4:17b-scout` | `mlx_lm.generate --model mlx-community/Llama-4-Scout-17B-16E-Instruct-4bit --prompt 'hello'` |
| llama-4-maverick | 401.58 (17) | 283.8 | 9.3 | 1327 | llama4 | `ollama run llama4:17b-maverick` | `mlx_lm.generate --model mlx-community/Llama-4-Maverick-17B-128E-Instruct-4bit --prompt 'hello'` |
| llama-3.3-70b | 70.55 | 51.4 | 7.7 | 1318 | llama3.3 | `ollama run llama3.3:70b` | `mlx_lm.generate --model mlx-community/Llama-3.3-70B-Instruct-4bit --prompt 'hello'` |
| llama-3.1-8b | 8.03 | 7.6 | 6.9 | 1211 | llama3.1 | `ollama run llama3.1:8b` | `mlx_lm.generate --model mlx-community/Llama-3.1-8B-Instruct-4bit --prompt 'hello'` |
| llama-3.2-3b | 3.21 | 4.3 | 5.7 | 1167 | llama3.2 | `ollama run llama3.2:3b` | `mlx_lm.generate --model mlx-community/Llama-3.2-3B-Instruct-4bit --prompt 'hello'` |
| mistral-large-4 | 1050 (52) | 726.5 (est.) | - | 1432 | unknown | - | - |
| mistral-large-3 | 675 (41) | 470 | 9.3 | 1414 | apache-2.0 | - | `mlx_lm.generate --model mlx-community/Mistral-Large-3-675B-Instruct-2512-mlx-4bit --prompt 'hello'` |
| mistral-medium-3.5-128b | 127.7 | 94 | 14.2 | 1426 | other | `ollama run mistral-medium-3.5:128b` | `mlx_lm.generate --model mlx-community/Mistral-Medium-3.5-128B-4bit --prompt 'hello'` |
| mistral-small-4-119b | 119.4 (6.5) | 85 | 11.3 | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/Mistral-Small-4-119B-2603-4bit --prompt 'hello'` |
| mistral-small-3.2-24b | 24.01 | 19.2 | 7 | 1356 | apache-2.0 | `ollama run mistral-small3.2:24b` | `mlx_lm.generate --model mlx-community/Mistral-Small-3.2-24B-Instruct-2506-4bit --prompt 'hello'` |
| magistral-small-2509 | 24.01 | 18.1 | 8.6 | - | apache-2.0 | `ollama run magistral:24b` | `mlx_lm.generate --model lmstudio-community/Magistral-Small-2509-MLX-4bit --prompt 'hello'` |
| devstral-small-2-24b | 24.01 | 19.2 | 7.5 | - | apache-2.0 | `ollama run devstral-small-2:24b` | `mlx_lm.generate --model mlx-community/Devstral-Small-2-24B-Instruct-2512-4bit --prompt 'hello'` |
| devstral-2-123b | 125.03 | 88.2 | 8.6 | - | other | `ollama run devstral-2:123b` | `mlx_lm.generate --model mlx-community/Devstral-2-123B-Instruct-2512-4bit --prompt 'hello'` |
| ministral-3-3b | 3.85 | 5.4 | 4.8 | - | apache-2.0 | `ollama run ministral-3:3b` | `mlx_lm.generate --model mlx-community/Ministral-3-3B-Instruct-2512-4bit --prompt 'hello'` |
| ministral-3-8b | 8.92 | 8.9 | 5.5 | - | apache-2.0 | `ollama run ministral-3:8b` | `mlx_lm.generate --model mlx-community/Ministral-3-8B-Instruct-2512-4bit --prompt 'hello'` |
| ministral-3-14b | 13.95 | 12.5 | 6 | - | apache-2.0 | `ollama run ministral-3:14b` | `mlx_lm.generate --model mlx-community/Ministral-3-14B-Instruct-2512-4bit --prompt 'hello'` |
| deepseek-v4.1-flash | 763.21 | 637.8 | 39.5 | 1475 | mit | - | `mlx_lm.generate --model mlx-community/DeepSeek-V4.1-Flash-MLX-4bit --prompt 'hello'` |
| deepseek-v4-pro | 1650.5 | 1140.8 (est.) | 36 | 1465 | mit | - | - |
| deepseek-v3.2 | 685.36 | 468.2 | 21.5 | 1425 | mit | - | `mlx_lm.generate --model mlx-community/DeepSeek-V3.2-4bit --prompt 'hello'` |
| deepseek-r1-0528-qwen3-8b | 8.19 | 8 | 8.1 | - | mit | `ollama run deepseek-r1:8b` | `mlx_lm.generate --model mlx-community/DeepSeek-R1-0528-Qwen3-8B-4bit --prompt 'hello'` |
| deepseek-ocr | 3.34 | 4.8 | - | - | mit | `ollama run deepseek-ocr:3b` | `mlx_lm.generate --model mlx-community/DeepSeek-OCR-4bit --prompt 'hello'` |
| phi-4-14b | 14.66 | 12.5 | 5.9 | 1256 | mit | `ollama run phi4:14b` | `mlx_lm.generate --model mlx-community/phi-4-4bit --prompt 'hello'` |
| phi-4-mini | 3.84 | 4.9 | 6.3 | - | mit | `ollama run phi4-mini:3.8b` | `mlx_lm.generate --model mlx-community/Phi-4-mini-instruct-4bit --prompt 'hello'` |
| phi-4-reasoning-plus | 14.66 | 14.6 | - | - | mit | `ollama run phi4-reasoning:14b-plus` | `mlx_lm.generate --model mlx-community/Phi-4-reasoning-plus-4bit --prompt 'hello'` |
| glm-5.3 | 753.33 | 483.1 | 44.8 | 1479 | glm-5.3 | - | `mlx_lm.generate --model mlx-community/GLM-5.3-4bit --prompt 'hello'` |
| glm-5.3-flash | 321.32 (18) | 233 | 41.8 | 1474 | mit | - | `mlx_lm.generate --model mlx-community/GLM-5.3-Flash-4bit --prompt 'hello'` |
| glm-4.7-flash | 31.22 (3) | 23.8 | 14.9 | 1365 | mit | `ollama run glm-4.7-flash` | `mlx_lm.generate --model mlx-community/GLM-4.7-Flash-4bit --prompt 'hello'` |
| glm-4.5-air | 110.47 (12) | 85.9 | 11.1 | 1373 | mit | - | `mlx_lm.generate --model mlx-community/GLM-4.5-Air-4bit --prompt 'hello'` |
| glm-ocr | 1.33 | 5.1 | - | - | mit | `ollama run glm-ocr` | - |
| kimi-k3 | 2779.93 | 1920.2 (est.) | 43.6 | 1487 | kimi-k3 | - | - |
| kimi-k2.7-code | 1026.88 | 739.2 | 25.8 | - | modified-mit | - | `mlx_lm.generate --model mlx-community/Kimi-K2.7-Code-4bit --prompt 'hello'` |
| kimi-k2.6 | 1026.88 | 710.5 (est.) | 27 | 1461 | modified-mit | - | - |
| kimi-linear-48b-a3b | 49.12 (3) | 36.6 | 7.3 | - | mit | - | `mlx_lm.generate --model mlx-community/Kimi-Linear-48B-A3B-Instruct-4bit --prompt 'hello'` |
| gpt-oss-20b | 20.91 (3.6) | 15.4 | 10 | 1317 | apache-2.0 | `ollama run gpt-oss:20b` | `mlx_lm.generate --model mlx-community/gpt-oss-20b-OptiQ-4bit --prompt 'hello'` |
| gpt-oss-120b | 116.83 (5.1) | 74.2 | 11.6 | 1352 | apache-2.0 | `ollama run gpt-oss:120b` | `mlx_lm.generate --model mlx-community/gpt-oss-120b-4bit --prompt 'hello'` |
| granite-4.2-3b | 3.66 | 4.5 | 9.1 | 1290 | apache-2.0 | `ollama run granite4.2:3b` | `mlx_lm.generate --model lmstudio-community/granite-4.2-3b-MLX-4bit --prompt 'hello'` |
| granite-4.2-8b | 8.79 | 8.1 | 11.1 | 1290 | apache-2.0 | `ollama run granite4.2:8b` | `mlx_lm.generate --model lmstudio-community/granite-4.2-8b-MLX-4bit --prompt 'hello'` |
| granite-4.2-30b | 29.28 | 22.7 | 12.7 | 1340 | apache-2.0 | `ollama run granite4.2:30b` | `mlx_lm.generate --model lmstudio-community/granite-4.2-30b-MLX-4bit --prompt 'hello'` |
| granite-embedding-278m | 0.28 | 2.3 | - | - | apache-2.0 | `ollama run granite-embedding:278m` | - |
| nemotron-3.5-super-120b | 124.45 (12) | 105.5 | 26.5 | 1416 | openmdw-1.1 | `ollama run nemotron-3.5-super:120b-a12b` | - |
| nemotron-3.5-bolt-74b | 74 (9) | 30.7 | 21.4 | - | openmdw-1.1 | `ollama run nemotron-3.5-bolt:74b-a9b` | - |
| nemotron-3.5-lightning-30b | 31.58 (3) | 30.7 | 12.9 | 1351 | openmdw-1.1 | `ollama run nemotron-3.5-lightning:30b-a3b` | - |
| nemotron-3-super-120b | 123.61 (12) | 102 | 12.8 | 1360 | nvidia-nemotron-open-model-license | `ollama run nemotron-3-super:120b-a12b` | - |
| nemotron-3-nano-30b | 31.58 (3) | 29.6 | 8.9 | 1314 | nvidia-nemotron-open-model-license | `ollama run nemotron-3-nano:30b-a3b` | - |
| nemotron-3-nano-4b | 3.97 | 5.2 | 4.8 | - | nvidia-nemotron-open-model-license | `ollama run nemotron-3-nano:4b` | - |
| nemotron-cascade-2-30b | 30 (3) | 29.6 | 11.7 | - | unknown | `ollama run nemotron-cascade-2:30b-a3b` | - |
| nemotron-3-ultra-550b | 560.52 (55) | 388.8 (est.) | 22.9 | 1426 | openmdw-1.1 | - | - |
| nemotron-3-nano-omni-30b | 33.02 (3) | 34.2 | 10.3 | - | nvidia-open-model-agreement | `ollama run nemotron3:33b` | - |
| olmo-3.1-32b-think | 32.23 | 23.8 | 7.1 | 1287 | apache-2.0 | `ollama run olmo-3.1:32b-think` | `mlx_lm.generate --model mlx-community/Olmo-3.1-32B-Think-4bit --prompt 'hello'` |
| olmo-3.1-32b-instruct | 32.23 | 23.8 | 6.5 | 1329 | apache-2.0 | `ollama run olmo-3.1:32b-instruct` | `mlx_lm.generate --model mlx-community/Olmo-3.1-32B-Instruct-4bit --prompt 'hello'` |
| olmo-3-7b-instruct | 7.3 | 7.2 | 5.6 | - | apache-2.0 | `ollama run olmo-3:7b-instruct` | `mlx_lm.generate --model mlx-community/Olmo-3-7B-Instruct-4bit --prompt 'hello'` |
| smollm3-3b | 3.08 | 4.2 | - | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/SmolLM3-3B-4bit --prompt 'hello'` |
| smollm2-1.7b | 1.71 | 3.3 | - | 1115 | apache-2.0 | `ollama run smollm2:1.7b-instruct` | - |
| minimax-m3 | 427.04 (23) | 302.5 | 29.2 | 1441 | minimax-community | - | `mlx_lm.generate --model mlx-community/MiniMax-M3-4bit --prompt 'hello'` |
| minimax-m2.7 | 228.69 | 161.1 | 22.8 | 1415 | other | - | `mlx_lm.generate --model mlx-community/MiniMax-M2.7-4bit --prompt 'hello'` |
| hermes-4.3-36b | 36.15 | 27 | - | - | apache-2.0 | - | - |
| hermes-4-14b | 14 | 12.3 | - | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/Hermes-4-14B-4bit --prompt 'hello'` |
| hermes-4-70b | 70.55 | 50.9 | 7.9 | - | llama3 | - | `mlx_lm.generate --model mlx-community/Hermes-4-70B-4bit --prompt 'hello'` |
| hermes-3-8b | 8.03 | 7.6 | - | - | llama3 | `ollama run hermes3:8b-llama3.1` | `mlx_lm.generate --model mlx-community/Hermes-3-Llama-3.1-8B-4bit --prompt 'hello'` |
| nouscoder-14b | 14.77 | 12.3 | - | - | apache-2.0 | - | - |
| laguna-xs-2.1 | 33.44 (3) | 25 | - | - | openmdw-1.1 | `ollama run laguna-xs-2.1` | `mlx_lm.generate --model mlx-community/Laguna-XS-2.1-4bit --prompt 'hello'` |
| laguna-s-2.1 | 117.56 (8) | 112.4 | - | - | openmdw-1.1 | `ollama run laguna-s-2.1` | - |
| north-mini-code-1.0 | 30.48 (3) | 23.8 | 9.9 | - | apache-2.0 | `ollama run north-mini-code-1.0` | `mlx_lm.generate --model mlx-community/North-Mini-Code-1.0-4bit --prompt 'hello'` |
| ornith-1.5-9b | 9 | 8.8 | - | - | mit | `ollama run ornith-1.5:9b` | `mlx_lm.generate --model mlx-community/Ornith-1.5-9B-OptiQ-4bit --prompt 'hello'` |
| ornith-1.5-35b | 35 (3) | 27.1 | - | - | mit | `ollama run ornith-1.5:35b` | `mlx_lm.generate --model mlx-community/Ornith-1.5-35B-A3B-OptiQ-4bit --prompt 'hello'` |
| lfm2.5-8b-a1b | 8.47 (1) | 8 | 7.2 | - | lfm1.0 | `ollama run lfm2.5:8b-a1b` | `mlx_lm.generate --model mlx-community/LFM2.5-8B-A1B-MLX-4bit --prompt 'hello'` |
| lfm2.5-24b-a2b | 23.97 (2.45) | 19.2 | - | - | lfm1.0 | `ollama run lfm2.5:24b` | `mlx_lm.generate --model LiquidAI/LFM2.5-24B-A2B-MLX-4bit --prompt 'hello'` |
| lfm2.5-1.2b-thinking | 1.17 | 2.8 | 5.2 | - | lfm1.0 | `ollama run lfm2.5-thinking:1.2b` | `mlx_lm.generate --model mlx-community/LFM2.5-1.2B-Thinking-4bit --prompt 'hello'` |
| minicpm-v-4.6 | 1.3 | 3.8 | 5.7 | - | apache-2.0 | `ollama run minicpm-v4.6` | - |
| minicpm-v-4.5 | 8.7 | 9 | - | - | apache-2.0 | `ollama run minicpm-v4.5` | - |
| minicpm5-2b | 2.52 | 3.9 | 12.5 | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/MiniCPM5-2B-mlx-4Bit --prompt 'hello'` |
| mimo-v2.6-flash | 310.76 (15) | 216.4 (est.) | 37.9 | 1451 | mit | - | - |
| mimo-v2.6-pro | 1024.22 (42) | 708.7 (est.) | 46.3 | 1479 | mit | - | - |
| k2-horizon-7b | 9 | 8.6 | 20.6 | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/K2-Horizon-7B-4bit --prompt 'hello'` |
| k2-horizon-mova-36b-a4b | 36 (4) | 28.1 | 25.3 | - | apache-2.0 | - | `mlx_lm.generate --model mlx-community/K2-Horizon-MoVA-36B-A4B-4bit --prompt 'hello'` |
| nomic-embed-text-v1.5 | 0.14 | 2.3 | - | - | apache-2.0 | `ollama run nomic-embed-text:v1.5` | - |
| nomic-embed-text-v2-moe | 0.48 | 3.1 | - | - | apache-2.0 | `ollama run nomic-embed-text-v2-moe` | - |
| mxbai-embed-large | 0.34 | 2.8 | - | - | apache-2.0 | `ollama run mxbai-embed-large:335m` | - |
| bge-m3 | 0.568 | 2.4 | - | - | mit | `ollama run bge-m3:567m` | `mlx_lm.generate --model mlx-community/bge-m3-mlx-4bit --prompt 'hello'` |
| astabrief-8b | 8.19 | 7.8 | - | - | apache-2.0 | - | - |
| kolibri-1 | 78.1 (3.46) | 56.6 | - | - | apache-2.0 | - | - |
| beam-1.0-501b-a23b | 501.27 (23.4) | 347.9 (est.) | - | - | apache-2.0 | - | - |
| mellum2.1-12b-a2.5b | 12.15 (2.5) | 11.3 | - | - | apache-2.0 | - | - |
| youtu-parsing-omni | 5.33 | 5.7 (est.) | - | - | youtu-parsing | - | - |
| vllm-sr-d3-27b | 26.09 | 24.3 | - | - | apache-2.0 | - | - |
| userrepgen-4b | 4.54 | 5.1 | - | - | mit | - | - |
| apertus-1.5-8b | 8.9 | 7.8 | - | - | apache-2.0 | - | `mlx_lm.generate --model tokimoa/apertus-v1.5-8b-mlx-4bit --prompt 'hello'` |
| apertus-1.5-70b | 72.01 | 52.3 | - | - | apache-2.0 | - | - |

## Sources

- Ollama library: https://ollama.com/library
- Hugging Face Hub API: https://huggingface.co/api/models
- Artificial Analysis leaderboard: https://artificialanalysis.ai/leaderboards/models
- LMArena Text leaderboard: https://lmarena.ai/leaderboard/text