hunyuan-t1-vision-20250619
A version of Tencent Hunyuan T1-Vision for multimodal understanding and deep thinking, supporting native multimodal long chain-of-thought reasoning. Tencent stated in the launch announcement that this…
MODEL DIRECTORY
New releases, in one place. Explore models by their publication date.
A version of Tencent Hunyuan T1-Vision for multimodal understanding and deep thinking, supporting native multimodal long chain-of-thought reasoning. Tencent stated in the launch announcement that this…
MiniMax Hailuo 02 is a video generation model released by MiniMax for creators and developers. It supports generating short videos from text or images. The official announcement emphasizes its ability…
Google's preview model for real-time interaction via the Gemini Live API, focused on low-latency, bidirectional voice and video interaction. (aistudio.google.com) The model is used for real-time…
A preview version of Gemini 2.5 Flash-Lite, intended for tasks sensitive to cost, latency, and throughput. At launch, Google positioned it as the lowest-cost, lowest-latency model in the Gemini 2.5…
Gemini 2.5 Flash is Google's multimodal reasoning model for low-latency, high-throughput tasks. It focuses on balancing performance and cost and is suitable for large-scale processing and agentic…
Gemini 2.5 Pro is Google's multimodal thinking model for complex reasoning, coding, and agentic tasks. Google announced its general release and availability as a stable version on June 17, 2025; the…
MiniMax-M1-80k is a long-context reasoning model in the MiniMax-M1 series, designed for complex tasks requiring long inputs and deep thinking. The model uses a mixture-of-experts (MoE) architecture…
MiniMax-M1-40k is a large-scale hybrid-attention reasoning model in the MiniMax M1 series. It uses a mixture-of-experts (MoE) and Lightning Attention architecture and is designed for complex tasks…
Kimi-Dev-72B is an open-source code large language model for software engineering problem solving. It is based on Qwen2.5-72B and has been trained on code-related data and optimized through…
Qianfan-QI-VL is a vision large model for image quality inspection developed in-house by Baidu AI Cloud's Qianfan platform. It is available as a preconfigured service and is primarily intended for…
A text-to-image preview version of Imagen 4 Ultra, focused on generating images that closely match text prompts. Input is a text prompt, and output is an image. (docs.cloud.google.com) Can generate…
Imagen 4 Fast Generate Preview is a public preview version of Imagen 4 fast image generation offered by Google on Vertex AI. Google announced the preview model ID in its release notes on June 11, 2025…
A preview version of the standard text-to-image model in Google's Imagen 4 series. It takes text prompts as input and outputs images; it is distinct from the Fast and Ultra versions.…
Doubao-Seed-1.6-flash is the high-speed version of the Doubao 1.6 model series, positioned for low-latency multimodal understanding and reasoning. The official release table states that it supports…
Doubao-Seed-1.6-thinking is a multimodal model in the Doubao Seed 1.6 series with enhanced deep thinking and visual understanding. The official Volcano Ark model release table lists…
Seedance 1.0 is a foundation model for video generation from ByteDance's Seed team. It was officially released on 2025-06-11, and its technical report was made public on the same day. (seed.bytedance.…
SeedEdit 3.0 is an instruction-based image editing model released by ByteDance's Seed team. Based on the text-to-image model Seedream 3.0, it generates edited results using an original image and text…
A translation-specific model from Tencent Hunyuan. This version is based on the Hunyuan 2B-Dense model, with translation capabilities optimized using multilingual supervised fine-tuning data, and uses…
Released by Google on June 5, 2025, the upgraded preview of Gemini 2.5 Pro is a multimodal thinking model for complex reasoning and programming tasks. (blog.google) Excels at code generation, complex…
A multilingual general-purpose text embedding model trained by Tongyi Lab based on Qwen3, belonging to the Qwen3-Embedding series. It converts text into numerical vectors rather than generating…