local.aiEarly access
HardwareModelsEnergyWallReferralsContact

Search

Find people, models, hardware, or pages

Account
HardwareModelsEnergyWallReferralsContactClaim your nameAccount
Model

Bonsai 27B

Intelligence rank
#77/238
53
Speed rank
#84/238
4m 48s
Fastest hardware: NVIDIA RTX PRO 6000 Blackwell 96GB.

Quick Start

Variations

Intelligence / Disk Size

Variations

Verified model-weight disk size. Higher intelligence and lower disk size are better.

best41475318.7GBDisk Size (lower →)Intelligence (higher →)awq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit · 53% intelligence · 18.7GB disk sizeawq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bitawq-4bit · prism-ml/Bonsai-27B-AWQ-4bit · 41% intelligence · 18.7GB disk sizeawq-4bit · prism-ml/Bonsai-27B-AWQ-4bit
Intelligence Density

Intelligence ÷ Disk Size (GB).

  1. 2.84
    awq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit
  2. 2.20
    awq-4bit · prism-ml/Bonsai-27B-AWQ-4bit
Hardware

Cost / task speed

Hardware cost / task speed
2 of 2 variations

Model variations

·
4153
Any intelligence
best4m 48s13m 03s35m 29s$16.3k$8.7k$4.7kHardware cost (lower →)Task time (faster →)awq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit (53) · NVIDIA DGX Spark 128GB · vllm · $4.7k · 30m 43sSparkawq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit (53) · NVIDIA GeForce RTX 5090 32GB · vllm · $6.4k · 4m 57sRTX 5090awq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit (53) · NVIDIA RTX 6000 Ada 48GB · vllm · $8.2k · 8m 16sRTX6k Adaawq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit (53) · NVIDIA RTX PRO 6000 Blackwell 96GB · vllm · $16.3k · 4m 48sRTXPro6kawq-4bit · prism-ml/Ternary-Bonsai-27B-AWQ-4bit (53)awq-4bit · prism-ml/Bonsai-27B-AWQ-4bit (41) · NVIDIA DGX Spark 128GB · vllm · $4.7k · 35m 29sSparkawq-4bit · prism-ml/Bonsai-27B-AWQ-4bit (41) · NVIDIA GeForce RTX 5090 32GB · vllm · $6.4k · 5m 42sRTX 5090awq-4bit · prism-ml/Bonsai-27B-AWQ-4bit (41) · NVIDIA RTX 6000 Ada 48GB · vllm · $8.2k · 9m 34sRTX6k Adaawq-4bit · prism-ml/Bonsai-27B-AWQ-4bit (41) · NVIDIA RTX PRO 6000 Blackwell 96GB · vllm · $16.3k · 5m 34sRTXPro6kawq-4bit · prism-ml/Bonsai-27B-AWQ-4bit (41)

Each coloured line is that variation’s cost / task-speed Pareto frontier. Global is the frontier across every qualifying variation; every point is still one original measured configuration.

Frontier →the ranked leaderboardAll models →every measured modelHardware →every measured device
local.aiIndependent local-AI benchmarksLive benchmark data
ContactAccountReferralsAbout