Add DeepSeek-V3 and DeepSeek-R1 to Open Foundation Models

[alvinreal]

Category 14: Open Models & Weights

Research completed for elite-tier open foundation models.

Projects Added (2)

Project	Stars	License	Description
DeepSeek-V3	103,430	MIT	671B parameter MoE model (37B activated) with state-of-the-art performance matching leading closed-source models. Features 128K context window, advanced reasoning capabilities, and efficient inference with FP8 support.
DeepSeek-R1	92,020	MIT	First-generation reasoning model with chain-of-thought capabilities. Open-weight model trained with large-scale reinforcement learning, achieving strong performance on math, code, and reasoning benchmarks.

Verification Criteria

• ✅ 1000+ GitHub stars (well above threshold)
• ✅ Active development within last 3 months
• ✅ OSI-approved open source licenses (MIT)
• ✅ Production-ready with comprehensive documentation

Notes

Both DeepSeek models represent significant contributions to the open-source AI ecosystem, offering competitive performance with closed-source alternatives while maintaining full open weights and permissive licensing.