AMD and NVIDIA Top Open Model Releases in 2026, Surpassing Major AI Labs

According to Hugging Face's State of Open Models: Summer 2026 report, AMD and NVIDIA each released over 200 new open-weight model repositories in 2026, outpacing every major AI lab including Meta, Google, and Alibaba. The chip makers are using freely released models as proof-of-concept demonstrations that their hardware can handle large-scale AI workloads, effectively treating open models as a hardware sales strategy. AMD has focused on converting trillion-parameter models to run on its own hardware, while NVIDIA has released models under its Nemotron family, including the 561B Nemotron 3 Ultra. Separately, Alibaba's Qwen has overtaken Meta's Llama as the community's default base model, accumulating over 151,000 derivative repositories — more than 2.6 times Meta's total and growing at 180–210 new derivatives per day. The report attributes Qwen's dominance to its consistent release cadence, wide size range, and permissive Apache 2.0 licensing.