LLM Providers
Deeployd supports multiple LLM providers out of the box, enabling you to choose the best model for each use case. All providers use a unified interface, making it easy to switch between models or use different providers for different agents.
Supported Providers
| Provider | Environment Variable | Official SDK |
|---|---|---|
| Anthropic (Claude) | ANTHROPIC_API_KEY | @anthropic-ai/sdk |
| OpenAI (GPT) | OPENAI_API_KEY | openai |
| Google (Gemini) | GOOGLE_API_KEY | Google AI |
| Mistral | MISTRAL_API_KEY | Mistral AI |
| Groq | GROQ_API_KEY | Groq |
Configuration
Set the environment variables for the providers you want to use:
# Required: At least one provider
ANTHROPIC_API_KEY=sk-ant-...
# Optional: Additional providers
OPENAI_API_KEY=sk-...
GOOGLE_API_KEY=AIza...
MISTRAL_API_KEY=...
GROQ_API_KEY=gsk_...
Multi-Provider Strategy
You don't need to configure all providers. Deeployd will automatically detect which providers are available based on the API keys present in your environment.
Available Models
Anthropic (Claude)
| Model ID | Name | Context | Max Output | Input Cost | Output Cost |
|---|---|---|---|---|---|
claude-opus-4-5-20251101 | Claude Opus 4.5 | 200K | 32K | $0.005/1K | $0.025/1K |
claude-sonnet-4-5-20250929 | Claude Sonnet 4.5 | 200K | 16K | $0.003/1K | $0.015/1K |
claude-haiku-4-5-20251015 | Claude Haiku 4.5 | 200K | 8K | $0.001/1K | $0.005/1K |
claude-3-5-sonnet-20241022 | Claude 3.5 Sonnet | 200K | 8K | $0.003/1K | $0.015/1K |
claude-3-5-haiku-20241022 | Claude 3.5 Haiku | 200K | 8K | $0.001/1K | $0.005/1K |
OpenAI (GPT)
| Model ID | Name | Context | Max Output | Input Cost | Output Cost |
|---|---|---|---|---|---|
gpt-4o | GPT-4o | 128K | 16K | $0.0025/1K | $0.01/1K |
gpt-4o-mini | GPT-4o Mini | 128K | 16K | $0.00015/1K | $0.0006/1K |
gpt-4-turbo | GPT-4 Turbo | 128K | 4K | $0.01/1K | $0.03/1K |
o1-preview | o1 Preview | 128K | 32K | $0.015/1K | $0.06/1K |
o1-mini | o1 Mini | 128K | 65K | $0.003/1K | $0.012/1K |
Google (Gemini)
| Model ID | Name | Context | Max Output | Input Cost | Output Cost |
|---|---|---|---|---|---|
gemini-2.0-flash | Gemini 2.0 Flash | 1M | 8K | $0.000075/1K | $0.0003/1K |
gemini-1.5-pro | Gemini 1.5 Pro | 2M | 8K | $0.00125/1K | $0.005/1K |
gemini-1.5-flash | Gemini 1.5 Flash | 1M | 8K | $0.000075/1K | $0.0003/1K |
Mistral
| Model ID | Name | Context | Max Output | Input Cost | Output Cost |
|---|---|---|---|---|---|
mistral-large-latest | Mistral Large | 128K | 8K | $0.002/1K | $0.006/1K |
mistral-medium-latest | Mistral Medium | 32K | 8K | $0.0027/1K | $0.0081/1K |
mistral-small-latest | Mistral Small | 32K | 8K | $0.0002/1K | $0.0006/1K |
codestral-latest | Codestral | 32K | 8K | $0.001/1K | $0.003/1K |
Groq (Fast Inference)
| Model ID | Name | Context | Max Output | Input Cost | Output Cost |
|---|---|---|---|---|---|
llama-3.3-70b-versatile | Llama 3.3 70B | 128K | 32K | $0.00059/1K | $0.00079/1K |
llama-3.1-70b-versatile | Llama 3.1 70B | 128K | 32K | $0.00059/1K | $0.00079/1K |
llama-3.1-8b-instant | Llama 3.1 8B | 128K | 8K | $0.00005/1K | $0.00008/1K |
mixtral-8x7b-32768 | Mixtral 8x7B | 32K | 8K | $0.00024/1K | $0.00024/1K |
Using Models in Agents
When creating an agent, specify the model ID:
const agent = await deeployd.agents.create({
name: 'My Agent',
model: 'claude-sonnet-4-5-20250929', // or 'gpt-4o', 'gemini-1.5-pro', etc.
systemPrompt: 'You are a helpful assistant.',
});
Model Selection Best Practices
By Use Case
| Use Case | Recommended Models |
|---|---|
| Complex reasoning | claude-opus-4-5, gpt-4o, o1-preview |
| General tasks | claude-sonnet-4-5, gpt-4o, gemini-1.5-pro |
| Fast responses | claude-haiku-4-5, gpt-4o-mini, gemini-2.0-flash |
| Code generation | claude-sonnet-4-5, codestral-latest, gpt-4o |
| Long context | gemini-1.5-pro (2M), claude-* (200K) |
| Cost-sensitive | gpt-4o-mini, gemini-2.0-flash, llama-3.1-8b-instant |
| Ultra-fast inference | Groq models (llama-*, mixtral-*) |
Model Capabilities
| Capability | Claude | GPT | Gemini | Mistral | Groq |
|---|---|---|---|---|---|
| Tool Use | Yes | Yes | Yes | Yes | Yes |
| Vision | Yes | Yes | Yes | No | No |
| Streaming | Yes | Yes | Yes | Yes | Yes |
| JSON Mode | Yes | Yes | Yes | Yes | Yes |
Model Orchestration
Deeployd includes intelligent model orchestration that can automatically:
- Route tasks to the most appropriate model based on complexity
- Fallback to alternative providers if one is unavailable
- Track performance across models for cost optimization
import { getModelOrchestrator } from '@deeployd/sdk';
const orchestrator = getModelOrchestrator();
// Automatic routing based on task type
const result = await orchestrator.execute({
taskType: 'code-generation',
messages: [{ role: 'user', content: 'Write a sorting function' }],
});
Cost Management
Deeployd tracks token usage and costs across all providers. View cost breakdowns in the dashboard or via API:
const usage = await deeployd.usage.get({
startDate: '2025-01-01',
endDate: '2025-01-31',
});
console.log(usage.byModel); // Cost breakdown by model
console.log(usage.byAgent); // Cost breakdown by agent
Getting API Keys
| Provider | Get Your Key |
|---|---|
| Anthropic | console.anthropic.com |
| OpenAI | platform.openai.com/api-keys |
| aistudio.google.com | |
| Mistral | console.mistral.ai |
| Groq | console.groq.com |