Verified Open Catalog
The Open AI Ecosystem Directory
Continuously updated index of state-of-the-art foundation models, inference engines, evaluation benchmarks, and distributed training tools.
| Resource | Type | Architecture / Specs | License | Primary Strengths |
|---|---|---|---|---|
| DeepSeek-V3 / R1 | Open Weights Foundation | 671B MoE (37B active parameters per token) | MIT Open Weights | SOTA reasoning, math, coding at 95% lower training cost |
| Meta Llama 3.3 / 4 | Open Weights Foundation | 70B Dense / MoE Variants | Llama Community | Enterprise standard, rich fine-tuning ecosystem |
| Qwen 2.5 Coder | Code Specialized LLM | 0.5B to 32B Dense | Apache 2.0 | Exceptional local coding, HumanEval 92%+ |
| vLLM | Inference Engine | PagedAttention, Tensor Parallelism | Apache 2.0 | Industry standard high-throughput serving |
| Ollama | Local Runtime | llama.cpp GGUF C++ backend | MIT | Zero-config local model execution on Mac/Windows/Linux |
| SWE-bench Verified | Evaluation Framework | 500 human-validated GitHub issues | MIT | The gold standard for autonomous software engineering |