OpenLuxOne API for every major model
OpenLux exposes text, image, audio, video and embedding models from 24 providers through a single OpenAI-compatible endpoint — one key, one bill and real-time usage details. Rates below are pay as you go and follow official provider pricing.
Every OpenLux model and its rate
Filter by modality to compare names, providers, discounts and the current pay-as-you-go rate for each model.
gpt-6-astra
-96%GPT-6 Astra is the latest flagship chat model, built for complex reasoning, agents and tool calling.
gpt-5.6-sol
-96%GPT-5.6 Sol is the value tier of the gpt-5.6 family, balancing quality, speed and cost.
gemini-3.8-flash
-92%Gemini 3.8 Flash handles general workloads with stable long-context, high-throughput calls.
Preview of Gemini 3 Flash — a low-cost option for high-concurrency workloads.
Claude Sonnet 4.5 is well balanced across coding, long context and tool use.
claude-opus-5
-91%Claude Opus 5 is Anthropic's flagship model for the hardest reasoning tasks.
claude-fable-5-1
-82%Claude Fable 5.1 targets long-form creation and research with consistently high-quality output.
Preview of Gemini 3.1 Pro with multimodal input and deep thinking modes.
deepseek-v4
-90%DeepSeek V4 offers strong Chinese comprehension and reasoning for regional workloads.
qwen3-max
-88%Qwen3 Max delivers dependable Chinese and coding performance.
All rates are shown in USD. Model availability, context length and final rates follow the console. For enterprise pricing or private deployment,contact us.
Which model should you use?
Reasoning, chat and agents
gpt-6-astra and Claude Opus 5 for the hardest reasoning, Claude Sonnet 4.5 and gpt-5.6-sol for balanced production traffic, and Gemini 3.8 Flash, DeepSeek V4 or Qwen3 Max when throughput matters more than depth.
Image, video, music and speech
FLUX 2 Pro for photoreal images and editing, Kling 3.0 for text-to-video with camera control, Suno V6 for full songs, and Doubao Voice 2 for low-latency multilingual speech.
Retrieval and RAG
BGE-M3 for multilingual embeddings and Jina Reranker v3 to lift retrieval accuracy before the answer model runs.
How OpenLux model pricing works
Every model is metered pay as you go: token-based models are billed per million input and output tokens, request-based models are billed per call, and the rate shown above is the rate applied to your account. Rates track official provider pricing, so a price change upstream is reflected in the catalogue and in the console rather than hidden in a multiplier.
Monthly and yearly plans add a token-equivalent credit allowance instead of locking you to one vendor, and the console shows cost per request in real time. Use the LLM API cost calculator to turn representative usage into a monthly estimate. Volume and enterprise agreements are handled through sales. See pricing plans for the credit allowances and the developer docs for the endpoints and error codes you will call.
Model questions
Do I need a separate API key for each provider?
No. Every model in the catalogue is called with the same OpenLux key against the same OpenAI-compatible endpoint. Switching from OpenAI to Claude, Gemini or DeepSeek means changing the model field, nothing else.
Which models can I call through OpenLux?
Text and reasoning models such as gpt-6-astra, gpt-5.6-sol, Claude Sonnet 4.5, Claude Opus 5, Gemini 3.8 Flash, DeepSeek V4 and Qwen3 Max; image, video and music models including FLUX 2 Pro, Kling 3.0 and Suno V6; and embedding models such as BGE-M3 and Jina Reranker v3.
How is model usage priced?
Usage is metered pay as you go in USD, per token or per request, at rates that follow official provider pricing — the rate and discount for each model are listed above. Monthly and yearly plans add a token-equivalent credit allowance on top.
How often are new models added?
Models from Western and Chinese providers are added as they become available upstream. The catalogue, the release notices and the model list in llms.txt are updated together.