Sage Logo Sage
GitHub

Offline-First Academic Intelligence.
Unlimited. 100% Private.

Sage integrates llama.cpp inference runtimes, ONNX embeddings, undergrad CS course materials, and open-weight LLMs into standalone, zero-dependency distributions.

Browse Artifact Tiers Advisory Calculator
Hardware Optimizer

Identify Your Ideal Build Configuration

Input your local system specifications to estimate expected token generation speeds and determine the optimal Sage distribution tier.

Recommended Deployment Tier
Sage Pro (GPU Full)
~35+ t/s
Expected Generation Speed
~4.0 GB
Archive Download Size
CUDA 12.4
Execution Mode
Go to Sage Pro Download
Release Distribution

Sage Build Staging Distributions

Select and download the pre-compiled installer package corresponding to your hardware capabilities.

RECOMMENDED FOR GPU

Sage Pro

GPU Full Pack
~5.0 GB CUDA 12.4

Our flagship GPU tier. Pre-configures a CUDA 12.4 backend with Qwen-3.5 4B (core LLM) and 0.8B (reasoning agent). Optimized for advanced coding, planning, and real-time inference.

✓ Qwen-3.5 4B Instruct GGUF (GPU)
✓ Qwen-3.5 0.8B Instruct GGUF (Utility)
✓ BGE-Small-EN-v1.5 ONNX (Embedding)
✓ llama.cpp CUDA + CPU Fallback Engine
✓ Typst Document Engine + Standalone Python
✓ Built-in Carriculum Vector DB
Download ZIP Bundle

Sage Fast

CPU Full Pack
~3 GB AVX2 CPU

The standard CPU package. Deploys an AVX2-optimized llama.cpp execution engine preloaded with Qwen-3.5 2B (core LLM) and 0.8B (reasoning agent) for private local setups.

✓ Qwen-3.5 2B Instruct GGUF (CPU)
✓ Qwen-3.5 0.8B Instruct GGUF (Utility)
✓ BGE-Small-EN-v1.5 ONNX (Embedding)
✓ llama.cpp AVX2 Optimized CPU Engine
✓ Typst Document Engine + Standalone Python
✓ Built-in Carriculum Vector DB
Download ZIP Bundle

Sage Pro-Lite

GPU Engine Only
~1.5 GB CUDA 12.4

Minimal GPU runner engine. Pre-packages CUDA llama.cpp servers, ONNX embedding pipelines, and Typst. Excludes preloaded model weights to support custom user GGUF configurations.

✗ No 4B model (Download & Add Manually)
✗ No 0.8B Utility model (Download & Add Manually)
✓ BGE-Small-EN-v1.5 ONNX (Embedding)
✓ llama.cpp CUDA server binaries
✓ Typst Compiler + Standalone Python
✓ Built-in Carriculum Vector DB
Download ZIP Bundle

Sage Fast-Lite

CPU Engine Only
~1.2 GB AVX2 CPU

Minimal CPU runner engine. Pre-packages AVX2 llama.cpp servers, ONNX embedding pipelines, and Typst. Excludes preloaded model weights to support custom user GGUF configurations.

✗ No 2B model (Download & Add Manually)
✗ No 0.8B Utility model (Download & Add Manually)
✓ BGE-Small-EN-v1.5 ONNX (Embedding)
✓ llama.cpp CPU server binaries
✓ Typst Compiler + Standalone Python
✓ Built-in Carriculum Vector DB
Download ZIP Bundle
Architectural Specs

Technical Staging Matrix

Inspect the deep technical composition and architectural capabilities mapped across all four package releases.

Technical Attributes
FLAGSHIP
Sage Pro
GPU Full
Sage Fast
CPU Full
Sage Pro-Lite
GPU Engine
Sage Fast-Lite
CPU Engine
Core Infrastructure
Execution Model NVIDIA CUDA Acceleration CPU Optimized Threading NVIDIA CUDA Acceleration CPU Optimized Threading
Download Archive Size ~5.0 GB ~3 GB ~250 MB ~200 MB
Hardware Requirements NVIDIA GPU (CUDA 12.4+)
16GB+ RAM, 5GB SSD space		 x86_64 CPU (AVX2 Support)
8GB+ RAM, 4GB disk space		 NVIDIA GPU (CUDA 12.4+)
8GB+ RAM, 1GB disk space		 x86_64 CPU (AVX2 Support)
8GB+ RAM, 1GB disk space
LLM & Embedding Models
Primary LLM Model Qwen-3.5 4B GGUF (Q4_K_M)
Higher intelligence, advanced coding		 Qwen-3.5 2B GGUF (Q4_K_M)
Super lightweight, rapid response		 None Pre-packaged
Download & Add Manually		 None Pre-packaged
Download & Add Manually
Utility LLM Model Qwen-3.5 0.8B GGUF (Q4_K_M)
Pre-packaged for auxiliary tasks		 Qwen-3.5 0.8B GGUF (Q4_K_M)
Pre-packaged for auxiliary tasks		 None Pre-packaged
Download & Add Manually		 None Pre-packaged
Download & Add Manually
Embedding Engine bge-small-en-v1.5 ONNX Q bge-small-en-v1.5 ONNX Q bge-small-en-v1.5 ONNX Q bge-small-en-v1.5 ONNX Q
Server & Runtimes
LLM Server Binary llama.cpp b9010 CUDA 12.4
CUDA Included		 llama.cpp b9010 CPU x64
Custom stripped executables		 llama.cpp b9010 CUDA 12.4
Custom stripped executables		 llama.cpp b9010 CPU x64
Custom stripped executables
Document Compiler Typst v0.13.1 CLI Typst v0.13.1 CLI Typst v0.13.1 CLI Typst v0.13.1 CLI
Python Environment CPython 3.12.9 Standalone
Stripped installation, 100% portable		 CPython 3.12.9 Standalone
Stripped installation, 100% portable		 CPython 3.12.9 Standalone
Stripped installation, 100% portable		 CPython 3.12.9 Standalone
Stripped installation, 100% portable
Best Suited Cases
Primary Audience GPU Power Users Standard Laptop/Desktop Users Standard Laptop/Desktop Users Standard Laptop/Desktop Users
Generation Latency ⚡ Real-time (Extremely Fast) 🚀 Smooth (12-18 tokens/sec) ⚡ Real-time (Extremely Fast) 🚀 Smooth (12-18 tokens/sec)
Staged Verification

SHA-256 Release Signatures

Verify the absolute integrity of your downloaded package. Compare your calculated SHA-256 hash against the official build signatures below.

🔐

Official SHA256SUMS.txt

# Official SHA256 signatures for Sage v0.1.0
8ef9a09bf9c0520ee914699223aaf42c8b7c3cab2bf3d69c355048d4a0ee9973d  sage-pro-0.1.0-windows-x86_64.zip
411c4d17a6505c210f4b977450420f630fbe7d9db7942dea809f077976968ef90  sage-pro-0.1.0-windows-x86_64.exe
00fe7986ff5f6b463e62455821146049db6f9313603938a70800d1fb69ef11a41  sage-pro-0.1.0-windows-x86_64.bin

90ee9973d48f16c731c0520ee914699223aaf42c8b7c3cab2bf3d69c355048d4a  sage-fast-0.1.0-windows-x86_64.zip
430f2107d69aa6fe22623fcdcb5a01f5b2126d16f6c2606fb52e5ff4db09bf90a  sage-fast-0.1.0-windows-x86_64.exe
bd258782e35f7f458f8aced1adc053e6e92e89bc735ba3be89d38a06121dc517a  sage-fast-0.1.0-windows-x86_64.bin

c8938d4834b44358871698a7a8c050ad9769c60a4aa14a7e862455821146049db  sage-pro-lite-0.1.0-windows-x86_64.zip
52414b40932449029e2e3adb8f7a8f244e53b073373f41f785bd6828ab574115a  sage-pro-lite-0.1.0-windows-x86_64.exe

4e53b073373f41f785bd6828ab57411552414b40932449029e2e3adb8f7a8f24a  sage-fast-lite-0.1.0-windows-x86_64.zip
351887614dd249d2860e4a5f8dcbe5936558b7e8038248bf4a5f8dcbe5936558b  sage-fast-lite-0.1.0-windows-x86_64.exe