> ## Documentation Index
> Fetch the complete documentation index at: https://docs.weventures.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# AI & LLM Providers
> Language models, AI frameworks, and orchestration tools
## Major LLM Providers
### OpenAI
**GPT-5, GPT-5 Mini, GPT-5 Nano, o3, DALL-E, Whisper**
Industry leader in large language models. GPT-5 for coding & agents, o3 for complex reasoning.
**Best for:** Production AI apps, coding, agentic tasks, embeddings
**Models (2025):**
* **GPT-5** - State-of-the-art coding & agents (74.9% SWE-bench, 88% Aider)
* **GPT-5 Mini** - Balanced performance & cost
* **GPT-5 Nano** - Ultra-fast & affordable
* **GPT-5 Pro** - Highest quality with scaled reasoning
* **o3** - Advanced reasoning model
* **DALL-E 3** - Image generation
* **Whisper** - Speech-to-text
**Context:** 272K input tokens, 128K output tokens
**Pricing (per 1M tokens):**
* GPT-5: $1.25 input / $10 output
* GPT-5 Mini: $0.25 input / $2 output
* GPT-5 Nano: $0.05 input / $0.40 output
* Prompt caching: \$0.125/1M tokens
**Key Features:**
* 45% fewer factual errors with web search
* 80% fewer hallucinations vs o3
* 4 reasoning levels: minimal, low, medium, high
* Parallel & sequential tool calling
* Available to free users
[Documentation](https://platform.openai.com/docs) • [API](https://openai.com/api)
### Anthropic
**Claude Sonnet 4.5, Opus 4.1, Haiku 4**
Best coding model in the world. Long context (1M tokens), autonomous agents, safety-focused.
**Best for:** Coding, long-running autonomous tasks, safety-critical apps
**Models (2025):**
* **Claude Sonnet 4.5** - Best coding model (77.2% SWE-bench, 61.4% OSWorld)
* **Claude Opus 4.1** - Most capable for complex reasoning
* **Claude Haiku 4** - Fastest & most affordable
**Context:** Up to 1M tokens (with beta header)
**Pricing (per 1M tokens):**
* Sonnet 4.5: $3 input / $15 output
* Opus 4.1: $15 input / $75 output
* 90% savings with prompt caching
* 50% savings with batch processing
**Key Features:**
* 64K output tokens for rich code generation
* Runs autonomously for 30+ hours
* State-of-the-art cybersecurity & compliance
* More resistant to prompt injection
* Outperforms Opus 4.1 in most areas
[Documentation](https://docs.anthropic.com) • [API](https://anthropic.com)
### Google AI
**Gemini 2.5 Pro, Flash**
Advanced reasoning model with thinking capabilities. Native audio, 1M+ context.
**Best for:** Complex reasoning, multimodal apps, cost-effectiveness
**Models (2025):**
* **Gemini 2.5 Pro** - State-of-the-art reasoning (18.8% Humanity's Last Exam)
* **Gemini 2.5 Pro Deep Think** - Extended reasoning variant
* **Gemini 2.0 Flash** - Ultra-fast & affordable
**Context:** 1M tokens (2M coming soon)
**Pricing (per 1M tokens):**
* Standard (\<200K): $1.25 input / $10 output
* Long context (>200K): $2.50 input / $15 output
* Flash: $0.10 input / $0.40 output
**Key Features:**
* Native audio outputs (24 languages)
* Thinking tokens for reasoning
* 63.8% SWE-bench Verified
* \#1 on LMArena leaderboard
* Cheaper than Anthropic, competitive with OpenAI
[Documentation](https://ai.google.dev/docs) • [API](https://ai.google.dev)
### Cohere
**Command, Embed, Rerank**
Enterprise-focused LLMs with strong embeddings and RAG optimization.
**Best for:** Enterprise deployments, embeddings, retrieval-augmented generation
**Models:**
* **Command R+** - Most capable
* **Command R** - Balanced performance
* **Embed v3** - Best-in-class embeddings
* **Rerank** - Search result ranking
**Pricing:** \$\$\$ (enterprise pricing)
[Documentation](https://docs.cohere.com) • [API](https://cohere.com)
## Fast Inference Providers
**Ultra-fast LLM inference with LPU**
500+ tokens/sec, lowest latency in industry. Perfect for real-time applications.
**Speed:** Fastest (LPU technology)
**Models:** Llama 3.1, Mixtral, Gemma
**Best for:** Real-time AI, chatbots, low-latency requirements
**Pricing:** \$ (generous free tier)
[Get Started](https://groq.com)
**Fast inference + fine-tuning**
Open-source models with custom fine-tuning capabilities.
**Best for:** Custom models, fine-tuning, open-source LLMs
**Features:**
* Fine-tuning support
* Open-source models
* Competitive pricing
* Fast inference
**Pricing:** \$\$ (mid-range)
[Get Started](https://together.ai)
**Production-grade inference**
Fast inference for open models with function calling support.
**Best for:** Serverless AI, function calling, production scale
**Features:**
* Serverless deployment
* Function calling
* Open-source models
* Fast performance
**Pricing:** \$\$ (pay-per-use)
[Get Started](https://fireworks.ai)
## Aggregators & Multi-Model Platforms
**Unified API for 300+ models**
Access OpenAI, Anthropic, Google, Groq, Meta through one API with automatic fallbacks.
**Best for:** Multi-model apps, cost optimization, avoiding vendor lock-in
**Features:**
* 300+ models
* Automatic fallbacks
* Cost optimization
* Single API
**Pricing:** 5% markup on base costs
[Get Started](https://openrouter.ai)
**Run open-source models**
Stable Diffusion, LLaMA, Whisper, and more. Pay-per-use serverless.
**Best for:** Open-source models, image/video generation, experimentation
**Features:**
* Open-source models
* Image generation
* Video generation
* Serverless deployment
**Pricing:** \$ (pay-per-use)
[Get Started](https://replicate.com)
## AI Frameworks & Orchestration
**Framework for building LLM apps**
v1.0 alpha released! New create\_agent built on LangGraph runtime.
**Best for:** RAG, chatbots, complex LLM workflows
**Latest Features (2025):**
* LangChain 1.0 alpha with unified agent
* Built on LangGraph runtime
* Python & JavaScript support
* Enhanced observability in LangSmith
* Open Agent Platform (no-code builder)
[Documentation](https://python.langchain.com)
**Build stateful AI agents (GA)**
Production-ready platform for deploying long-running autonomous agents.
**Best for:** Complex agents, multi-step workflows, production deployments
**Latest Features (2025):**
* LangGraph Platform (Generally Available)
* Node caching & deferred nodes
* Revision queueing for smooth deploys
* Trace mode with LangSmith integration
* Dynamic tool calling
* Pre/Post model hooks
* Built-in web search & RemoteMCP
* Studio v2 (runs locally)
* 1-click GitHub deployment
[Documentation](https://langchain-ai.github.io/langgraph)
**Multi-agent orchestration**
Role-based agents collaborating on tasks. Hierarchical or sequential workflows.
**Best for:** Multi-agent systems, collaborative AI, task delegation
* Multi-agent
* Role-based
* Collaborative workflows
* Easy setup
[Documentation](https://docs.crewai.com)
**TypeScript AI framework**
Streaming responses, function calling, React hooks. Built for Next.js.
**Best for:** Next.js apps, edge AI, streaming UIs, React integration
* TypeScript-first
* Streaming support
* React hooks
* Edge compatible
[Documentation](https://sdk.vercel.ai/docs)
**Data framework for LLMs**
Data connectors, indexes, query engines. Optimized for RAG.
**Best for:** RAG applications, data ingestion, structured retrieval
* Data connectors
* RAG-optimized
* Query engines
* Indexing strategies
[Documentation](https://docs.llamaindex.ai)
**Multi-agent framework (Microsoft)**
Conversable agents with human-in-the-loop and code execution.
**Best for:** Research, multi-agent collaboration, code generation
* Multi-agent
* Code execution
* Human-in-loop
* Microsoft Research
[Documentation](https://microsoft.github.io/autogen)
## Provider Comparison (2025)
| Provider | Speed | Cost | Context | Best For |
| ------------------------- | --------- | ------ | ------- | ------------------------------------- |
| **OpenAI GPT-5** | Fast | \$\$ | 272K | Coding, agents, production AI |
| **Anthropic Sonnet 4.5** | Medium | \$\$\$ | 1M | Best coding, autonomous agents |
| **Google Gemini 2.5 Pro** | Medium | \$\$ | 1M-2M | Reasoning, multimodal, cost-effective |
| **Groq** | ⚡ Fastest | \$ | Varies | Real-time, low latency |
| **OpenRouter** | Fast | \$\$ | Varies | Multi-model flexibility |
| **Together AI** | Fast | \$\$ | Varies | Open models, fine-tuning |
## Pricing Comparison (per 1M tokens - 2025)
**Input Tokens:**
* GPT-5: \$1.25
* GPT-5 Mini: \$0.25
* GPT-5 Nano: \$0.05
* Claude Sonnet 4.5: \$3.00
* Claude Opus 4.1: \$15.00
* Gemini 2.5 Pro (standard): \$1.25
* Gemini 2.5 Pro (long context): \$2.50
* Gemini 2.0 Flash: \$0.10
**Output Tokens:**
* GPT-5: \$10.00
* GPT-5 Mini: \$2.00
* GPT-5 Nano: \$0.40
* Claude Sonnet 4.5: \$15.00
* Claude Opus 4.1: \$75.00
* Gemini 2.5 Pro (standard): \$10.00
* Gemini 2.5 Pro (long context): \$15.00
* Gemini 2.0 Flash: \$0.40
## Edge Functions vs Custom SDKs
**Deno-based serverless functions**
**Pros:**
* Direct database access
* Global deployment
* TypeScript/JavaScript
* Built-in secrets management
* Free tier included
**Cons:**
* Deno runtime only
* Supabase ecosystem
**Best for:** Supabase users, TypeScript developers, database-connected AI
```typescript theme={null}
// Supabase Edge Function example
import { serve } from "https://deno.land/std@0.168.0/http/server.ts"
serve(async (req) => {
const { prompt } = await req.json()
const response = await fetch("https://api.openai.com/v1/chat/completions", {
method: "POST",
headers: {
"Authorization": `Bearer ${Deno.env.get("OPENAI_API_KEY")}`,
"Content-Type": "application/json"
},
body: JSON.stringify({
model: "gpt-4",
messages: [{ role: "user", content: prompt }]
})
})
return new Response(JSON.stringify(await response.json()))
})
```
**Official OpenAI, Anthropic SDKs**
**Pros:**
* Official support
* Latest features
* Type-safe
* Well-documented
* Works anywhere
**Cons:**
* More boilerplate
* Need separate hosting
* API key management
**Best for:** Any runtime, maximum flexibility, latest features
```typescript theme={null}
// OpenAI SDK example
import OpenAI from "openai"
const openai = new OpenAI({
apiKey: process.env.OPENAI_API_KEY
})
const completion = await openai.chat.completions.create({
model: "gpt-4",
messages: [{ role: "user", content: "Hello!" }]
})
```
## Memory & State Management
**Lightweight React state**
Perfect for managing AI conversation state in React apps.
* Minimal boilerplate
* React integration
* TypeScript support
* Persistent storage
[Documentation](https://zustand-demo.pmnd.rs)
**Built-in conversation memory**
Buffer, summary, entity, and vector store memory types.
* Conversation history
* Summary memory
* Entity extraction
* Vector store integration
[Documentation](https://python.langchain.com/docs/modules/memory)
**Serverless Redis for state**
Perfect for storing conversation history and session data.
* Serverless
* Low latency
* Global replication
* REST API
[Documentation](https://upstash.com/docs/redis)
**Edge-compatible key-value store**
Built on Upstash, perfect for Next.js edge functions.
* Edge-compatible
* Low latency
* Simple API
* Vercel integration
[Documentation](https://vercel.com/docs/storage/vercel-kv)
## Decision Trees
**Need industry-leading performance?** → OpenAI (GPT-4)
**Long documents & safety-critical?** → Anthropic (Claude)
**Real-time, low-latency chatbots?** → Groq
**Want flexibility across providers?** → OpenRouter
**Open-source models & fine-tuning?** → Together AI
**Image/video generation?** → Replicate
**Cost-effectiveness & multimodal?** → Google AI (Gemini)
**Building RAG applications?** → LangChain + LlamaIndex
**Complex multi-step agents?** → LangGraph
**Multi-agent collaboration?** → CrewAI
**Next.js with streaming UI?** → Vercel AI SDK
**Research & experimentation?** → AutoGen
**Simple chat integration?** → Direct SDK (OpenAI, Anthropic)
## Recommended Combinations
**OpenAI** (embeddings) + **Pinecone** (vectors) + **LangChain** (orchestration)
**Groq** (inference) + **Upstash Redis** (state) + **Vercel AI SDK** (streaming)
**OpenRouter** (multi-model) + **Chroma** (vectors) + **Supabase Edge Functions**
**Anthropic Claude** (safety) + **Weaviate** (hybrid search) + **LangGraph** (agents)