Online Supernode Rental
Online access to GPU Supernode resources with request-based provisioning, unified delivery, and room for future elastic expansion.
Production AI inference
on wylon
A high-performance LLM inference platform built for developers and enterprises.
Core products · Token
Factory
Token Factory(public sign-up coming soon)
Pay-per-token inference for leading open-source LLMs. OpenAI- and Anthropic-compatible APIs, ready to use with your existing clients.
Supported model families: MiniMax · Kimi · GLM · Qwen · DeepSeek
MiniMax
MiniMax
Long-context, productivity-grade workloads — document processing, summarization, multi-turn business conversations.
Best forLong-form /
structured
Use casesEnterprise knowledge /
long documents
Moonshot
Kimi
Solid on multimodal, ultra-long context, and code — a popular foundation for agents and developer tooling.
Best forLong dialogue /
repos
Use casesAgent / IDE /
research
Zhipu · Z.ai
GLM
Strong on Chinese corpora — covers general chat, tool use, and agentic workflows.
Best forGeneral chat / tool
calls
Use casesCustomer support /
Chinese-language workloads
Qwen
Qwen
Full lineup from on-device tiny models to flagship MoEs, balancing performance and cost.
Best forGeneral / multilingual /
multimodal
Use casesContent / RAG /
vision
DeepSeek
DeepSeek
Reputation for reasoning, code, and math — its MoE line offers excellent price-performance.
Best forReasoning / code /
chain-of-thought
Use casesCode / agent
planning
And more
See the full list, context lengths, and pricing in the model catalog and pricing pages.
wylon continuously adds new models — the authoritative, up-to-date list is the Model Plaza in your console.
Core products · GPU Supernode
GPU Supernode
GPU compute for online supernode rental and enterprise dedicated service. Now accepting early-access requests.
Two delivery options
Coming soon
Core technology
Inference foundation for domestic GPUs
Token Factory and GPU Supernode share the same technology foundation. Through an LLM inference OS that coordinates vertical storage resources and horizontal GPU fabric, wylon turns domestic GPU capacity from isolated resources into system-level inference infrastructure.
Products & Platform
One foundation, two product forms
Token Factory delivers model inference directly through APIs;
GPU Supernode delivers compute resources for inference infrastructure.
Developers and enterprises can choose by usage pattern and resource needs, while both share the same system foundation.
wylon Supernode System
Hiten OS coordinates vertical storage and horizontal compute to form a system-level LLM inference foundation for domestic GPUs.
Operating System Plane
System coordination for horizontal compute and vertical storage
Through unified scheduling, Hiten organizes vertically integrated storage and horizontally networked compute into one complete runtime space, allowing model workloads to translate more directly into hardware efficiency.
Storage Plane
Vertical integration keeps data flowing continuously
By connecting context, cache, and data-access layers, the storage plane provides sustained support for long-context and high-concurrency LLM inference.
Compute Plane
Horizontal GPU fabric turns single nodes into a coordinated system
High-speed interconnects and coordination between GPU units organize distributed resources into a horizontal compute plane for scalable parallel inference.
Extending the Pareto frontier of domestic GPUs
The wylon Supernode System improves single-user response speed while sustaining higher system throughput, extending the efficiency frontier for domestic GPUs in large-scale model inference scenarios.
Same speed, higher throughput; same throughput, faster response
Why wylon
Built for LLM inference workloads
Full-stack, hardware-software co-designed
End-to-end cloud infrastructure across GPU compatibility, super-node architecture, inference runtime, and API services — tightly integrated to reduce performance overhead.
Enterprise
Service quality
A high-availability architecture built for production workloads, delivering dependable, enterprise-grade service quality.
0
Ops overhead
Drop-in inference API — no infrastructure to manage.
10×
Speedup
Backed by system-level caching, prefill is up to 10× faster than baseline implementations.
6+
Chip families supported
wylon's super-node system supports Biren, MetaX, Sunrise, Cambricon, and more.
Silicon partners
Get started
Try the wylon Token Factory
Token Factory is open for sign-ups and can be integrated in minutes.