Kingy AI Launch Intelligence
Track new AI tools, AI agents, model releases, AI startups, funding announcements, AI video tools, AI coding tools, and generative AI product launches.
The Qwen team released Qwen3, a new generation of open-weight models with reasoning, coding, math, and general capability improvements.
A core open-model record because Qwen3 strengthened the non-US open-weight model ecosystem across reasoning and coding.
Meta introduced the Llama 4 herd with Scout and Maverick as open-weight natively multimodal mixture-of-experts models, plus a Behemoth preview.
Important for the hub because Llama 4 reset Meta's open-model direction around multimodal MoE models.
Mistral AI released Mistral Small 3.1, an Apache 2.0 multimodal and multilingual model with improved text performance and 128K context.
A useful open-model record because small capable models often matter more in production than headline frontier systems.
DeepSeek released DeepSeek-R1 and DeepSeek-R1-Zero, open reasoning models with distilled variants based on Qwen and Llama models.
A top-priority record because R1 changed the conversation around open reasoning, cost, and model competition.
DeepSeek released DeepSeek-V3, a large mixture-of-experts language model with open weights and a technical report available through GitHub and Hugging Face.
Important because DeepSeek-V3 became one of the most watched open model releases before R1 shocked the market.
Cohere released Command R7B, a compact R-series model focused on speed, efficiency, RAG, tool use, and agentic behavior.
A strong fit for the hub because small, capable models are often more practical than flagship models for real deployments.
Stability AI released Stable Diffusion 3.5 Large and Large Turbo as open model weights, followed by Stable Diffusion 3.5 Medium.
A core open-weight image model record because it affects builders who want more control than closed image APIs allow.
Meta released Llama 3.2 with 11B and 90B vision models plus lightweight 1B and 3B text models for edge and mobile devices.
A high-value open-model record for builders watching local and multimodal deployment.
The Qwen team released Qwen2.5-Coder as an open-source coding model family and renamed CodeQwen to Qwen-Coder.
A useful coding-model record because Qwen2.5-Coder helped expand serious open alternatives for code generation.
Mistral AI announced Pixtral 12B as a natively multimodal model for image and text understanding with a 128K context window.
Useful for tracking open multimodal model progress before the 2025 wave of stronger vision models.
Black Forest Labs launched and released the FLUX.1 suite of text-to-image models, including commercial and open-weight variants.
Important because FLUX.1 became a leading open-weight image model family and a real alternative to older Stable Diffusion workflows.
Meta launched Llama 3.1 including the 405B open-weight frontier model, upgraded 8B and 70B models, 128K context, multilingual support, and tool use.
Essential for the hub because Llama 3.1 was a serious open alternative to closed frontier models.
Mistral AI released Mistral NeMo, a 12B model built with NVIDIA, offering a 128K context window and Apache 2.0 checkpoints.
Important as an efficient open-weight alternative for teams that cannot host giant frontier models.
Stability AI released Stable Diffusion 3 Medium weights as a smaller open image generation model in the Stable Diffusion 3 series.
Worth tracking because open image model launches often have a second-order effect through fine-tunes, LoRAs, and creator tools.
Microsoft introduced Phi-3, a family of small language models aimed at efficient, capable AI on constrained hardware and lower-cost deployments.
A useful open-model record because Phi-3 helped normalize serious small-model deployment for practical apps.
Meta released Llama 3 with 8B and 70B pretrained and instruction-tuned language models, plus safety tools for developers.
Foundational open-weight record because Llama 3 became a default option for developers wanting capable deployable models.
Methodology
Launch profiles prioritize official sources, launch dates, categories, audience fit, pricing clarity, demo links, and last-verified dates before they are treated as indexable records.
Hub collections use structured signals such as Product Hunt activity, GitHub or Hugging Face links, funding notes, video demos, founder submissions, and Kingy editorial scores.
Scores favor launches with a clear use case, visible audience, understandable pricing, credible demo material, and enough context to support a useful article, comparison, or creator review.
We feature the best AI apps, tools, and platforms across the web. If you are an AI app creator and would like to be featured here, feel free to contact us.
