Open Source Models

Open-source AI models offer strong performance, often at lower costs. These models are developed by companies and research labs that share their work with the community.

Why Open Source?

Open-source models offer several advantages:

• Lower costs - Often cheaper to run
• Transparency - You can see how they work
• No vendor lock-in - Not tied to one company
• Community improvements - Continuous enhancements
• Self-hosting option - Run on your own infrastructure

Available Open-Source Models

Meta (Llama)

Llama 4 Maverick

When to use: Creative writing, casual conversations, general tasks.

Llama 4 Scout

When to use: Everyday tasks where cost matters.

Mistral

Mistral Large

When to use: Complex tasks, European languages, reasoning.

Pixtral Large

When to use: Analyzing images, visual tasks.

DeepSeek

DeepSeek V3.1

When to use: Code generation, logical reasoning, technical tasks.

DeepSeek R1

When to use: Math, logic problems, when you want to see step-by-step thinking.

Qwen

Qwen3 Coder

When to use: Writing code, especially in multiple programming languages.

Qwen3 235B

When to use: Tasks requiring maximum capability from open-source models.

xAI (Grok)

Grok 3

When to use: Current events, conversational tasks.

Grok 4

When to use: Complex reasoning, important tasks.

GLM (zAI)

GLM 4.5

When to use: Budget-friendly general use.

Choosing an Open-Source Model

Open Source vs. Proprietary

When to use open source:

• Cost is a priority
• You want good results without premium pricing
• You value transparency
• Specific strengths match your needs (e.g., DeepSeek for coding)

When to use proprietary:

• Quality is paramount
• You need the latest capabilities
• You want reliable, consistent performance
• Official support matters

Model Highlights

Best for Coding: DeepSeek V3.1 or Qwen3 Coder

Both excel at code generation and understanding. DeepSeek has stronger reasoning, while Qwen3 Coder is specialized for programming.

Example: "Write a Python function to merge two sorted lists"

Best for Creative Writing: Llama 4 Maverick

Produces engaging, creative content with good flow and style.

Example: "Write a short story opening about a detective in 1940s Los Angeles"

Best for Reasoning: DeepSeek R1

Shows step-by-step thinking, great for math and logic.

Example: "If all bloops are razzies and all razzies are lazzies, are all bloops lazzies?"

Best for Budget: GLM 4.5 or Llama 4 Scout

Solid performance at just 1 credit per message.

Common Questions

Next Steps

• Compare with OpenAI models
• Configure model settings
• Learn about credits