← Back to all models

gpt neox 20b

General model by eleutherai · 20.74B parameters

Chip
Unified Memory
Bandwidth 307 GB/s
Available for Models ~21 GB

Everything this chip runs, machine by machine: MacBook Pro 14" M5 Pro, MacBook Pro 16" M5 Pro

Runs Excellently

gpt neox 20b fits comfortably in your 24 GB Mac using Q6_K quantization, using 77% of your RAM.

~10 tok/s

Q6_K · 18.5 GB · Estimated

Quantization Options

QuantizationMemorySpeedFits?
Q8_0 23.6 GB~8 tok/s✗
Q6_K Recommended 18.5 GB~10 tok/s✓
Q5_K_M 16.0 GB~12 tok/s✓
Q4_K_M 13.9 GB~14 tok/s✓
Q3_K_M 11.8 GB~17 tok/s✓
Q2_K 9.5 GB~22 tok/s✓

Specifications

Parameters20.7B
Architecturegpt_neox
Context Length2K
CategoryGeneral
Formatgguf
Minimum RAM11.6 GB
Licenseapache-2.0
HuggingFace Downloads364,757

GGUF Sources

mradermacher
View on HuggingFace →

Which Mac Can Run gpt neox 20b?

Minimum Mac

MacBook Pro 14" — M5 Pro 24GB

Q6_K · ~10 tok/s · 77% RAM

MacQuanttok/sRAM %
Mac Studio · M3 Ultra 96GB Q8_0 ~21 25%
Mac Studio · M3 Ultra 256GB Q8_0 ~21 9%
Mac Studio · M3 Ultra 512GB Q8_0 ~21 5%
Mac Studio · M1 Ultra 64GB Q8_0 ~20 37%
Mac Studio · M1 Ultra 128GB Q8_0 ~20 18%
Mac Studio · M2 Ultra 64GB Q8_0 ~20 37%
Mac Studio · M2 Ultra 128GB Q8_0 ~20 18%
Mac Studio · M2 Ultra 192GB Q8_0 ~20 12%
MacBook Pro 14" · M5 Max 36GB Q8_0 ~16 66%
MacBook Pro 14" · M5 Max 48GB Q8_0 ~16 49%
MacBook Pro 14" · M5 Max 64GB Q8_0 ~16 37%
MacBook Pro 14" · M5 Max 128GB Q8_0 ~16 18%
MacBook Pro 16" · M5 Max 36GB Q8_0 ~16 66%
MacBook Pro 16" · M5 Max 48GB Q8_0 ~16 49%
MacBook Pro 16" · M5 Max 64GB Q8_0 ~16 37%
MacBook Pro 16" · M5 Max 128GB Q8_0 ~16 18%
MacBook Pro 14" · M4 Max 36GB Q8_0 ~14 66%
MacBook Pro 14" · M4 Max 48GB Q8_0 ~14 49%
MacBook Pro 14" · M4 Max 64GB Q8_0 ~14 37%
MacBook Pro 14" · M4 Max 128GB Q8_0 ~14 18%
MacBook Pro 16" · M4 Max 36GB Q8_0 ~14 66%
MacBook Pro 16" · M4 Max 48GB Q8_0 ~14 49%
MacBook Pro 16" · M4 Max 64GB Q8_0 ~14 37%
MacBook Pro 16" · M4 Max 128GB Q8_0 ~14 18%
Mac Studio · M4 Max 36GB Q8_0 ~14 66%
Mac Studio · M4 Max 48GB Q8_0 ~14 49%
Mac Studio · M4 Max 64GB Q8_0 ~14 37%
Mac Studio · M4 Max 128GB Q8_0 ~14 18%
MacBook Pro 14" · M5 Pro 24GB Q6_K ~10 77%
MacBook Pro 16" · M5 Pro 24GB Q6_K ~10 77%
MacBook Pro 14" · M1 Max 32GB Q8_0 ~10 74%
MacBook Pro 14" · M1 Max 64GB Q8_0 ~10 37%
MacBook Pro 14" · M2 Max 32GB Q8_0 ~10 74%
MacBook Pro 14" · M2 Max 64GB Q8_0 ~10 37%
MacBook Pro 14" · M2 Max 96GB Q8_0 ~10 25%
MacBook Pro 14" · M3 Max 36GB Q8_0 ~10 66%
MacBook Pro 14" · M3 Max 48GB Q8_0 ~10 49%
MacBook Pro 14" · M3 Max 64GB Q8_0 ~10 37%
MacBook Pro 14" · M3 Max 96GB Q8_0 ~10 25%
MacBook Pro 14" · M3 Max 128GB Q8_0 ~10 18%
MacBook Pro 16" · M1 Max 32GB Q8_0 ~10 74%
MacBook Pro 16" · M1 Max 64GB Q8_0 ~10 37%
MacBook Pro 16" · M2 Max 32GB Q8_0 ~10 74%
MacBook Pro 16" · M2 Max 64GB Q8_0 ~10 37%
MacBook Pro 16" · M2 Max 96GB Q8_0 ~10 25%
MacBook Pro 16" · M3 Max 36GB Q8_0 ~10 66%
MacBook Pro 16" · M3 Max 48GB Q8_0 ~10 49%
MacBook Pro 16" · M3 Max 64GB Q8_0 ~10 37%
MacBook Pro 16" · M3 Max 96GB Q8_0 ~10 25%
MacBook Pro 16" · M3 Max 128GB Q8_0 ~10 18%
Mac Studio · M1 Max 32GB Q8_0 ~10 74%
Mac Studio · M1 Max 64GB Q8_0 ~10 37%
Mac Studio · M2 Max 32GB Q8_0 ~10 74%
Mac Studio · M2 Max 64GB Q8_0 ~10 37%
Mac Studio · M2 Max 96GB Q8_0 ~10 25%
MacBook Pro 14" · M4 Pro 24GB Q6_K ~9 77%
MacBook Pro 16" · M4 Pro 24GB Q6_K ~9 77%
Mac mini · M4 Pro 24GB Q6_K ~9 77%
MacBook Pro 14" · M5 Pro 48GB Q8_0 ~8 49%
MacBook Pro 14" · M5 Pro 64GB Q8_0 ~8 37%
MacBook Pro 16" · M5 Pro 48GB Q8_0 ~8 49%
MacBook Pro 16" · M5 Pro 64GB Q8_0 ~8 37%
MacBook Pro 14" · M4 Pro 48GB Q8_0 ~7 49%
MacBook Pro 16" · M4 Pro 48GB Q8_0 ~7 49%
Mac mini · M4 Pro 48GB Q8_0 ~7 49%
Mac mini · M4 Pro 64GB Q8_0 ~7 37%
MacBook Air · M5 24GB Q6_K ~5 77%
MacBook Pro 14" · M5 24GB Q6_K ~5 77%
MacBook Pro 14" · M1 Pro 32GB Q8_0 ~5 74%
MacBook Pro 14" · M2 Pro 32GB Q8_0 ~5 74%
MacBook Pro 16" · M1 Pro 32GB Q8_0 ~5 74%
MacBook Pro 16" · M2 Pro 32GB Q8_0 ~5 74%
Mac mini · M2 Pro 32GB Q8_0 ~5 74%
MacBook Air · M4 24GB Q6_K ~4 77%
MacBook Pro 14" · M4 24GB Q6_K ~4 77%
Mac mini · M4 24GB Q6_K ~4 77%
iMac · M4 24GB Q6_K ~4 77%
MacBook Air · M5 32GB Q8_0 ~4 74%
MacBook Pro 14" · M5 32GB Q8_0 ~4 74%
MacBook Pro 14" · M3 Pro 36GB Q8_0 ~4 66%
MacBook Pro 16" · M3 Pro 36GB Q8_0 ~4 66%
MacBook Air · M2 24GB Q6_K ~3 77%
MacBook Air · M3 24GB Q6_K ~3 77%
MacBook Pro 14" · M3 24GB Q6_K ~3 77%
Mac mini · M2 24GB Q6_K ~3 77%
iMac · M3 24GB Q6_K ~3 77%
MacBook Air · M4 32GB Q8_0 ~3 74%
MacBook Pro 14" · M4 32GB Q8_0 ~3 74%
Mac mini · M4 32GB Q8_0 ~3 74%
iMac · M4 32GB Q8_0 ~3 74%
MacBook Pro 14" · M1 Pro 16GB Q3_K_M ~11 74%
MacBook Pro 14" · M2 Pro 16GB Q3_K_M ~11 74%
MacBook Pro 16" · M1 Pro 16GB Q3_K_M ~11 74%
MacBook Pro 16" · M2 Pro 16GB Q3_K_M ~11 74%
Mac mini · M2 Pro 16GB Q3_K_M ~11 74%
MacBook Air · M5 16GB Q3_K_M ~8 74%
MacBook Pro 14" · M5 16GB Q3_K_M ~8 74%
MacBook Pro 14" · M3 Pro 18GB Q4_K_M ~7 77%
MacBook Pro 16" · M3 Pro 18GB Q4_K_M ~7 77%
MacBook Air · M4 16GB Q3_K_M ~7 74%
MacBook Pro 14" · M4 16GB Q3_K_M ~7 74%
Mac mini · M4 16GB Q3_K_M ~7 74%
iMac · M4 16GB Q3_K_M ~7 74%
MacBook Air · M2 16GB Q3_K_M ~6 74%
MacBook Air · M3 16GB Q3_K_M ~6 74%
MacBook Pro 14" · M3 16GB Q3_K_M ~6 74%
Mac mini · M2 16GB Q3_K_M ~6 74%
iMac · M3 16GB Q3_K_M ~6 74%
MacBook Air · M1 16GB Q3_K_M ~4 74%
Mac mini · M1 16GB Q3_K_M ~4 74%
iMac · M1 16GB Q3_K_M ~4 74%
MacBook Air · M2 8GB — — 119%
MacBook Air · M3 8GB — — 119%
MacBook Pro 14" · M3 8GB — — 119%
Mac mini · M2 8GB — — 119%
iMac · M3 8GB — — 119%
MacBook Air · M1 8GB — — 119%
Mac mini · M1 8GB — — 119%
iMac · M1 8GB — — 119%

Models similar to gpt neox 20b

Closest matches by use case and parameter count.

Run gpt neox 20b locally on your Mac

ToolPiper downloads, manages, and runs models with one click. Apple Silicon optimized.

Get ToolPiper — Free