Select Categories
Select Developers
Select Model Types

DeepSeek-V3

Open Source Mixture-of-Experts

A state-of-the-art Mixture-of-Experts language model with 671B total parameters and 37B activated per token, delivering competitive performance with leading closed-source models.

DeepSeek
Type: Language Model
Parameters: 671B total, 37B activated

Best For:

Mathematical computationsCode developmentResearch applicationsLarge document analysisMulti-lingual tasksOpen source projects

GPT-OSS

Open Source Agentic AI

OpenAI's open-source GPT model series featuring configurable reasoning effort, full chain-of-thought capabilities, and agentic functions including web browsing and Python execution.

OpenAI
Type: Language Model
Parameters: 120B (5.1B active), 20B (3.6B active)

Best For:

Agentic AI applicationsResearch and developmentWeb-enabled AI assistantsCode execution tasksReasoning-intensive applicationsProduction AI systems

Kimi K2

Open Source Mixture-of-Experts Agentic AI

Moonshot AI's trillion-parameter mixture-of-experts model designed for open agentic intelligence, featuring advanced reasoning, tool use, and autonomous problem-solving capabilities.

Moonshot AI
Type: Language Model
Parameters: 1T total, 32B active

Best For:

Agentic AI systemsAutonomous workflowsResearch applicationsCustom AI solutionsMultilingual applicationsCost-effective deploymentsOpen source projects

Llama

Open Source Multimodal

Meta's most advanced generation of Llama models featuring natively multimodal capabilities, mixture-of-experts architecture, advanced reasoning, and industry-leading context windows.

Meta
Type: Language Model
Parameters: Scout, Maverick, Behemoth

Best For:

Enterprise applicationsLong document analysisMultimodal AI assistantsVisual reasoning tasksScalable AI deploymentCost-efficient AI solutions