JANGQ-AI/MiniMax-M2.7-JANG_3L

Text Generation • 26B • Updated 6 days ago • 3.4k • 12

Note 3-bit | 89 GB | 95.5% MMLU (200q) — Best value

JANGQ-AI/MiniMax-M2.7-JANG_6M

Text Generation • 47B • Updated 6 days ago • 1.08k

Note 6-bit | 167 GB | near-lossless

JANGQ-AI/MiniMax-M2.7-JANG_2L

Text Generation • 19B • Updated 6 days ago • 4.11k • 11

Note 2-bit | 63 GB | 88.0% MMLU (200q)

JANGQ-AI/Nemotron-Cascade-2-30B-A3B-JANG_2L

Text Generation • 3B • Updated 6 days ago • 681 • 3

Note JANG_2L — 10.3 GB — 130 tok/s — MMLU (no-think) 59.0% — MMLU (reasoning) 88.0% VS JANG_4M — 17 GB — MMLU (no-think) 69.0% — MMLU (reasoning) 93.0% VS MLX 4-bit — 16.6 GB — MMLU (no-think) 69.0% — MMLU (reasoning) 92.5% VS MLX 6-bit — 23.9 GB — MMLU (no-think) 71.0% — MMLU (reasoning) 94.5%

JANGQ-AI/Nemotron-3-Super-120B-A12B-JANG_4M

Text Generation • 18B • Updated 6 days ago • 856 • 1

Note JANG_4M — 63 GB — 55.1 tok/s — MMLU (no-think) 72.5% — MMLU (reasoning) 93.0% VS MLX 4-bit — 63 GB — MMLU (reasoning) 93.5% VS MLX 3-bit — CRASHES

JANGQ-AI/Qwen3.5-397B-A17B-JANG_2L

Text Generation • 54B • Updated 6 days ago • 539

Note JANG_2L — 187 GB — 36 tok/s — MMLU (no-think) 79.5% — MMLU (reasoning) 92.0% VS MLX 4-bit — 209 GB — MMLU (no-think) 81.5% — MMLU (reasoning) 94.0% VS MLX 2/3-bit — NaN (float16 overflow)

JANGQ-AI/Qwen3.5-122B-A10B-JANG_2S

11B • Updated 6 days ago • 512 • 1

Note JANG_2S (2-bit) - 35 GB - MMLU: 79% - 54 tok/s VS MLX 2-bit - 36 GB - MMLU: 56.5%

JANGQ-AI/Qwen3.5-397B-A17B-JANG_1L

Text Generation • 34B • Updated 6 days ago • 602 • 3

Note JANG_1L — 112 GB — 36.1 tok/s — MMLU (no-think) 81.0% — MMLU (reasoning) 86.5% — FITS 128 GB VS MLX 4-bit — 209 GB — MMLU (reasoning) 94.0% VS MLX 2/3-bit — NaN

JANGQ-AI/Qwen3.5-35B-A3B-JANG_4K

5B • Updated 6 days ago • 818 • 2

Note JANG_4K (4-bit K-quant) - 18 GB - MMLU: 77.5% VS MLX 4-bit - 18 GB - MMLU: 77.0%

JANGQ-AI/Qwen3.5-122B-A10B-JANG_4K

18B • Updated 6 days ago • 541 • 1

Note JANG_4K (4-bit K-quant) - 69 GB - MMLU: 86% - ~50 tok/s VS MLX 4-bit - 64 GB - MMLU: 85%

JANGQ-AI/Nemotron-Cascade-2-30B-A3B-JANG_4M

Text Generation • 5B • Updated 6 days ago • 537 • 2

Note JANG_4M — 17 GB — MMLU (no-think) 69.0% — MMLU (reasoning) 93.0% — BEATS MLX 4-bit (92.5%) VS JANG_2L — 10.3 GB — 130 tok/s — MMLU (reasoning) 88.0% — fits 16 GB Macs VS MLX 4-bit — 16.6 GB — MMLU (no-think) 69.0% — MMLU (reasoning) 92.5% VS MLX 6-bit — 23.9 GB — MMLU (reasoning) 94.5%

JANGQ-AI/Qwen3.5-4B-JANG_4S

1B • Updated 6 days ago • 511 • 2

Note JANG_4S (4-bit) - 3.0 GB - MMLU: 67.5% - 134 tok/s VS MLX 4-bit - 2.2 GB - MMLU: 67.0%

JANGQ-AI/Qwen3.5-35B-A3B-JANG_2S

3B • Updated 6 days ago • 539 • 1

Note JANG_2S (2-bit) - 11 GB - MMLU: 65.5% VS MLX 2-bit - 10 GB - MMLU: ~20%

JANGQ-AI/Qwen3.5-27B-JANG_4S

5B • Updated 6 days ago • 1.01k • 4

Note JANG_4S (4-bit) - 16 GB - MMLU: 84.5% - 35 tok/s VS MLX 4-bit - 14 GB - MMLU: 84.5% - 20 tok/s

JANGQ-AI/Qwen3.5-9B-JANG_4S

2B • Updated 6 days ago • 809 • 5

Note JANG_4S (4-bit) - 6.0 GB - MMLU: 73.0% VS MLX 4-bit - 4.7 GB - MMLU: 72.5%

JANGQ-AI/Nemotron-3-Super-120B-A12B-JANG_2L

Text Generation • 13B • Updated 6 days ago • 729 • 1

Note JANG_2L — 43 GB — 51.6 tok/s — MMLU (no-think) 75.0% — MMLU (reasoning) 86.0% VS JANG_4M — 63 GB — 55.1 tok/s — MMLU (reasoning) 93.0% VS MLX 4-bit — 63 GB — MMLU (reasoning) 93.5% VS MLX 3-bit — CRASHES

JANGQ-AI/Qwen3.5-122B-A10B-JANG_3L

14B • Updated 6 days ago • 536 • 1

Note JANG_3L (3-bit) - 49 GB - MMLU: 81.5% - 49.6 tok/s VS MLX 4-bit - 64 GB - MMLU: 85%

JANGQ-AI/MiniMax-M2.7-JANG_4M

Text Generation • 33B • Updated 6 days ago • 1.34k

Note 4-bit | 115 GB | 95.5% MMLU (200q)