Notifications

No notifications yet

Research Lab

Sunlight Artificial Intelligence

Yavqo Sunlight is our state-of-the-art suite of language models and neural architectures, engineered to execute securely across local operating systems and global cloud meshes.

Available Neural Models

Choose the specific capacity and parameter size that matches your latency and accuracy requirements.

Sunlight-L (Large)

Undisclosed (Frontier)

Our premier frontier model optimized for agentic workloads, multi-step reasoning, and long-context coding tasks.

Context Window:1,000,000 tokens
Throughput:120 tok/sec
API Pricing:$5.00 / M input | $30.00 / M output

Sunlight-M (Medium)

70B params

The industry-standard choice for real-time applications. High accuracy balanced with fast response times.

Context Window:64K tokens
Throughput:280 tok/sec
API Pricing:$0.35 / M tokens

Sunlight-S (Small)

8B params

Designed for offline use. Runs locally on Horizon OS devices with minimal battery drain and high privacy.

Context Window:16K tokens
Throughput:80 tok/sec (on-device)
API Pricing:Free (Local)
Custom Silicon

Yavqo Tensor Processing Grid

Our customized hardware accelerator chip, the TPG-1, powers local Sunlight-S models with absolute power efficiency, enabling secure offline intelligence.

  • 220 TFLOPS Neural Compute: Massively parallel pipelines for transformer operations.
  • 4.2W Active Thermal Envelope: Extremely low heat output, perfect for fanless Horizon OS laptops.
  • Dedicated Unified SRAM Cache: 64MB low-latency vector cache directly on the silicon core.
YAVQO TPG-1
Playground

Interact with Sunlight-L

Test Yavqo's leading LLM right here in the browser. Type in your prompt and see real-time streaming tokens.

Model Status: Active
sunlight-l
Sunlight AI
Welcome to the Yavqo Sunlight Playground. I am Sunlight-L, Yavqo's flagship frontier reasoning model. Ask me anything about our ecosystem: Horizon OS, Nebula Cloud, Forge IDE, or ask me a general coding question!