RedditDeepSeek V4 Official Release Scheduled for Mid-JulyJun 29, 20261 minDeepSeek announced official launch of V4 model for mid-July. Integration work already underway in llama.cpp ecosystem.
Reddit r/LocalLLaMANemotron-3-Super-120B: Hybrid Mamba+MoE Model with 504K Token RetrievalJun 27, 20261 minNemotron-3-Super-120B demonstrated perfect needle retrieval to 504K tokens using hybrid Mamba+MoE architecture. Significant long-context capability.
Chinese AI Lab GitHubDeepSeek-V3 GitHub Repository Active DevelopmentJun 27, 20261 minDeepSeek-V3 repository shows recent updates (2026-06-27) with 103,813 stars. Indicates active, substantial community adoption.
RedditNVIDIA Releases Nemotron-TwoTower-30B: Diffusion-Based Language ModelJun 25, 20261 minNVIDIA released Nemotron-TwoTower-30B-A3B-Base-BF16, an unusual diffusion-based language model built on Nemotron 3 architecture. Represents alternative approach to traditional autoregressive language modeling.
Reddit r/MachineLearningSoftmax-Free Attention Model with Triton Kernels at GPT-2 Medium ScaleJun 21, 20261 minNovel attention architecture achieving 354M parameters with long-context VRAM savings via structural sparsity and custom Triton kernels. Open weights released.
GitHubDeepSeek-R1: Major Update to DeepSeek Reasoning ModelJun 21, 20261 minDeepSeek-R1 repository shows significant activity with 91,986 stars. Represents latest iteration of Chinese open-source reasoning model gaining substantial adoption.
GitHubKimi-K2: Latest Model by Moonshot/KimiJun 20, 20261 minMoonshot's Kimi released Kimi-K2 with 10,866 GitHub stars. Active development with recent update 2026-06-20.
GitHubQwen3.6 foundation model released by AlibabaJun 19, 20261 minAlibaba releases Qwen3.6 model with 3,587 GitHub stars. Updated foundation model in open-source Qwen family.
GitHubQwen3-TTS text-to-speech model released by AlibabaJun 19, 20261 minAlibaba releases Qwen3-TTS with 12,029 GitHub stars. Multimodal capability expansion for Qwen model family.
GitHubQwen3-VL model reaches 19,406 GitHub starsJun 17, 20261 minAlibaba's Qwen3-VL multimodal model gains 19.4k+ GitHub stars. Updated 2026-06-17. Indicates strong adoption of vision-language model.