Glyph
Glyph is a research framework and vision-language model for long-context understanding. It renders long text as images and processes those images with a vision-language model, aiming to reduce…
MODEL DIRECTORY
New releases, in one place. Explore models by their publication date.
Glyph is a research framework and vision-language model for long-context understanding. It renders long text as images and processes those images with a vision-language model, aiming to reduce…
Seedance 1.0 Pro Fast is ByteDance’s video-generation model optimized for a balance of quality, speed, and cost. Volcengine says generation can be up to approximately three times faster than Seedance…
Seed3D 1.0 is a 3D generative foundation model designed to turn a single image into a high-fidelity, simulation-ready 3D asset. It was announced on October 23, 2025. (seed.bytedance.com) Generates 3D…
Qwen2.5-VL-Embedding is a unified multimodal embedding model based on Qwen2.5-VL. It converts text, images, video, or mixed-modality inputs into a shared representation vector. Alibaba lists it as…
Qwen3-Rerank is a text reranking model based on a Qwen LLM foundation, served through Alibaba Cloud Model Studio. It ranks candidate documents by their relevance to a query, supports more than 100…
DeepSeek-OCR is a vision-language model for optical character recognition and research into visual-text compression. DeepSeek announced its release on October 20, 2025. (github.com) Recognizes text…
Doubao-Seed-1.6-lite is a small-size conversational model positioned for deep reasoning. ByteDance states that its inference speed is substantially higher than that of doubao-seed-1-6-250615. The…
Doubao-Seed-1.6 is a ByteDance conversational model with deep-reasoning capabilities. The official release table identifies this model ID as a new October 2025 release supporting both thinking and non…
Claude Haiku 4.5 is a small model designed for low latency and cost efficiency. At launch, Anthropic reported coding performance comparable to Claude Sonnet 4 at one-third the cost and more than twice…
Veo 3.1 Fast Preview is Google’s speed-oriented preview model for generating short videos. Google announced it on October 15, 2025 as part of the Veo 3.1 preview family. (docs.cloud.google.com) The…
Veo 3.1 Preview is Google’s preview video-generation model for creating short clips from text and images, with synchronized audio. It was announced for Vertex AI on October 15, 2025. The standard…
Qwen3-VL-Flash is a small, fast-response vision-language model in the Qwen3 series. It combines thinking and non-thinking modes. The specified snapshot is qwen3-vl-flash-2025-10-15. (help.aliyun.com)…
Tencent’s Hunyuan translation model is a text-to-text model for multilingual translation. The announcement dated 2025-10-14 describes a model version update, not necessarily the model’s original…
Gemini 2.5 Computer Use is a specialized preview model built on Gemini 2.5 Pro’s visual understanding and reasoning capabilities. It was announced on October 7, 2025 for building agents that interact…
GPT Image 1 mini is a cost-efficient version of GPT Image 1. It is a natively multimodal model that accepts text and image inputs and produces images. It is intended for image generation and editing,…
GPT-Audio mini is a cost-efficient GPT Audio model for speech-to-speech applications. It accepts and produces audio and text, and is documented for use through the Chat Completions API. Potential uses…
GPT-Realtime mini is a cost-efficient version of GPT-Realtime for real-time speech-to-speech applications. OpenAI announced its release on October 6, 2025. (developers.openai.com) Accepts text, image…
Gemini 2.5 Flash Image is a multimodal model optimized for fast image understanding, generation, and editing, balancing cost and performance. (docs.cloud.google.com) Google announced general…
GLM-4.6 is a flagship language model positioned for coding, reasoning and agentic tasks. It was released on September 30, 2025. (docs.z.ai) Accepts text input and produces text output, with a…
DeepSeek-V3.2-Exp is an experimental large language model released on September 29, 2025. Built on DeepSeek-V3.1-Terminus, it introduces DeepSeek Sparse Attention (DSA) to improve training and…