Supported Providers
Lava supports 27 AI service providers out of the box, enabling you to route requests to dozens of different AI APIs through a single, unified billing system.
All providers are accessible through Lava’s /v1/forward endpoint with automatic usage tracking, billing, and cost calculation.
Provider Categories
Large Language Models (LLMs)
Chat completions, text generation, and conversational AI:
| Provider | Models | Key Features |
|---|
| OpenAI | GPT-4, GPT-3.5, GPT-4o | Chat, embeddings, streaming, function calling |
| Anthropic | Claude 3 (Opus, Sonnet, Haiku) | Long context (200K), vision, tool use |
| Google | Gemini Pro, Flash | Multimodal inputs, code generation |
| xAI | Grok | Real-time data, advanced reasoning |
| Mistral | Large, Medium, Small | Efficient inference, multilingual |
| DeepSeek | DeepSeek models | Cost-effective, high performance |
| Cohere | Command, Embed | Enterprise features, RAG support |
| Groq | Ultra-fast inference | Low latency, high throughput |
| together.ai | Open-source LLMs | Llama, Mixtral, custom models |
AI Infrastructure & Hosting
Platforms that host and serve AI models at scale:
- Fireworks, DeepInfra, Hyperbolic, Cerebras, SambaNova
- Nebius AI Studio, GMI Cloud, Inference.net, Baseten
Voice & Audio
Speech synthesis, recognition, and voice AI:
| Provider | Capabilities | Billing |
|---|
| ElevenLabs | Text-to-speech, speech-to-text, voice cloning | Character-based |
| Retell | AI phone calls, real-time responses | Duration-based |
- Vercel (AI SDK integration)
- Novita AI (model marketplace)
Specialized Services
- kluster.ai (workflow automation)
- Parasail, Chutes, Targon, Pearch
How Provider Integration Works
When you send a request through Lava:
- Automatic Routing: Lava identifies the provider from your target URL
- Authentication: Lava adds the appropriate provider API key
- Request Forwarding: Your request is sent to the provider unchanged
- Usage Tracking: Lava extracts usage metrics (tokens, characters, duration, etc.)
- Cost Calculation: Costs are calculated based on provider pricing + your configured fees
Lava handles provider-specific authentication formats, response parsing, and usage extraction automatically - you only need to provide the target URL.
Provider Authentication Methods
Lava automatically handles different authentication formats for each provider:
| Authentication Type | Providers | Header Format |
|---|
| Bearer Token | OpenAI, Anthropic, DeepSeek, Mistral, xAI, most LLMs | Authorization: Bearer <key> |
| x-api-key | Standard API key header | x-api-key: <key> |
| x-goog-api-key | Google Gemini | x-goog-api-key: <key> |
| xi-api-key | ElevenLabs | xi-api-key: <key> |
| Query Parameter | Google (alternative) | ?key=<key> |
You don’t need to manage provider API keys yourself - Lava uses managed keys for all providers. Your forward token handles authentication on the Lava side.
Metering by Provider Type
Token-Based Metering
Providers: OpenAI, Anthropic, Google, Mistral, xAI, Groq, Cohere, most LLMs
Metrics: Input tokens, output tokens, cached tokens, audio tokens
Pricing: Per million tokens (1M)
Character-Based Metering
Providers: ElevenLabs (text-to-speech)
Metrics: Character count of input text
Pricing: Per million characters (1M)
Duration-Based Metering
Providers: Retell (voice calls), ElevenLabs (speech-to-text)
Metrics: Audio duration, call duration
Pricing: Per minute or per second
Request-Based Metering
Providers: Image generation services
Metrics: Number of API requests
Pricing: Per request
Provider Availability
All providers are available on Lava’s production infrastructure with:
- ✅ Global edge deployment - Requests routed to nearest region
- ✅ Less than 20ms latency overhead - Minimal proxy delay
- ✅ 99.9% uptime SLA - Enterprise-grade reliability
- ✅ Automatic failover - Provider outages handled gracefully
- ✅ Real-time monitoring - Usage tracking and error logs
Next Steps
