⚠️ Important: The current Lightweight API is an internal project — originally built to solve our own needs. It is not a public service. Access is granted selectively to community members who request it. A fully public API (potentially under a different name or as a separate product) may be released in the future, but there is no timeline for this.
Lightweight is a unified AI API gateway that gives you access to 25+ models from OpenAI, Anthropic, Google, and xAI — through a single OpenAI-compatible endpoint. No vendor lock-in, no multiple API keys, no hassle.
# That's it. Any OpenAI-compatible tool works instantly.
export OPENAI_BASE_URL="https://api.lightweight.one/v1"
export OPENAI_API_KEY="lw_sk_your-key-here"| Provider | Models |
|---|---|
| OpenAI | GPT-5.4, GPT-5.4 Mini, GPT-5.2, GPT-5.1, GPT-5.1 Codex, GPT-5 Mini, GPT-4.1, GPT-4o, and more |
| Anthropic | Claude Opus 4.6, Claude Opus 4.6 Fast, Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Sonnet 4, Claude Opus 4.5, Haiku 4.5 |
| Gemini 3.1 Pro, Gemini 3 Pro, Gemini 3 Flash, Gemini 2.5 Pro | |
| xAI | Grok Code Fast |
More providers coming soon. We plan to expand to Meta (Llama), DeepSeek, Mistral, and others as backend support grows. Check
GET /v1/modelsfor the live catalog.
Claude Code, OpenCode, Cursor, Windsurf, Aider, Zed, Continue, Cline, LangChain, LiteLLM, and any OpenAI-compatible tool.
Models retain their full context windows through the gateway. For example, Claude Opus 4.6 supports the full 200K token context. Check GET /v1/models for each model's exact limits.
Note: Extended thinking (reasoning) is supported for Claude models. The model reasons internally and thinking blocks are returned in the API response.
Lightweight is currently in invite-only beta.
This project started as an internal solution to solve our own problems with AI API access. It's independent and experimental — not a commercial service with dedicated support teams or SLA guarantees.
Open a Discussion in the Access Requests category with:
- Who you are — brief introduction
- What you want to build — projects you plan to use the API for, or services you want to migrate
- Why Lightweight — what draws you to a unified gateway
We review requests and grant invite codes on a case-by-case basis.
| Aspect | Details |
|---|---|
| Support | Community-driven. No dedicated support team. Use Discussions and our documentation. |
| Uptime & SLA | Variable. We're independent and monitor actively, but we don't guarantee SLAs. |
| New Models | Frontier models from major providers typically arrive same-day within hours. Some models may take longer depending on availability. |
| Bugs & Issues | Expected in beta. Report via Issues. Community PRs welcome. |
| Pricing | Free during beta. No hidden costs. |
This is the community hub for Lightweight. Here's how to participate:
- Access Requests — Request an invite code
- General — Ask questions, share experiences
- Ideas & Suggestions — Propose features, integrations, improvements
- Show & Tell — Share what you've built with Lightweight
- Bug reports
- API behavior problems
- Model-specific issues
Found a bug? Have a fix? PRs are welcome. The community helps shape this project:
- Report issues you encounter
- Suggest improvements to docs or the API
- Share workarounds and tips in Discussions
- Help other users in the community
| Resource | Link |
|---|---|
| Documentation | docs.lightweight.one |
| API Base URL | https://api.lightweight.one/v1 |
| API Status | lightweight.one/status |
| Dashboard | lightweight.one/dashboard |
| Model Catalog | GET https://api.lightweight.one/v1/models |
Is this free? Yes, during the beta period. No credit card required.
Is this open source? The API itself is private. This repo is the public community hub for access requests, discussions, and issue tracking. The documentation is open source.
Can I trust this for production? Lightweight is experimental and independent. Uptime is variable. We recommend having a fallback for production-critical workloads.
How fast do new models arrive? Frontier models from major providers typically arrive same-day within hours of release. Some models may take longer depending on availability.
Which providers are supported? Currently OpenAI, Anthropic, Google, and xAI. We plan to expand to additional providers (Meta, DeepSeek, Mistral, etc.) in the future.
Is there dedicated support? No. The community supports itself through Discussions and documentation. We actively monitor but don't provide 1:1 support.