GPT-5.6 Terra
GPT-5.6 Terra is a GPT-5.6 family model designed to balance intelligence and cost. OpenAI describes it as roughly corresponding to the mini tier in earlier GPT-5 families. It was released on July 9, 2…
MODEL DIRECTORY
New releases, in one place. Explore models by their publication date.
GPT-5.6 Terra is a GPT-5.6 family model designed to balance intelligence and cost. OpenAI describes it as roughly corresponding to the mini tier in earlier GPT-5 families. It was released on July 9, 2…
GPT-5.6 Sol is OpenAI’s flagship GPT-5.6 model for complex professional work. OpenAI announced general availability on July 9, 2026, following a limited preview. It is intended for tasks including…
GPT-Live-1 mini is a voice model in OpenAI’s GPT-Live family, designed for natural, real-time conversations. At launch, OpenAI designated it as the default ChatGPT Voice model for Free users.…
GPT-Live-1 is OpenAI’s full-duplex voice model for natural, real-time conversations. It can listen and speak simultaneously, handle interruptions and pauses, and delegate deeper reasoning or tool use…
ByteDance announced Seedream 5.0 Pro on July 8, 2026 as a multimodal image-creation model aimed at complex, professional visual-design workflows. (seed.bytedance.com) Generates images from text or…
Release date: July 7, 2026, according to Tencent’s official announcement. (github.com) HunyuanOCR-1.5 is a lightweight, end-to-end vision-language model specialized in optical character recognition…
GPT-5.5 Instant Mini is a ChatGPT fallback model announced on July 6, 2026. At launch, it replaced GPT-5.3 Instant Mini for users who reached their GPT-5.5 Instant or Auto rate limits.…
GPT-Realtime-2.1 mini is a distilled reasoning model designed for faster, lower-cost realtime voice interactions. OpenAI announced its release on July 6, 2026. (developers.openai.com) Supports audio…
GPT-Realtime-2.1 is a realtime reasoning model for speech-to-speech interactions and voice-agent workflows. Compared with GPT-Realtime-2, OpenAI reports improvements in alphanumeric recognition,…
Hy3 is Tencent’s production language model for practical reasoning and agent workflows, released on July 6, 2026. It succeeds Hy3 preview. Tencent describes it as a hybrid fast-and-slow-thinking…
Alibaba’s Wan 2.7 reference-to-video model generates video using text and visual references, with support for audio timbre references. It is designed to improve consistency of characters, props and…
Wan 2.7 text-to-video is a video-generation model served through Alibaba Cloud Model Studio. Alibaba announced this version on July 1, 2026; the date in its ID identifies the June 12, 2026 snapshot,…
Qwen3-Rerank is a text reranking model trained on a Qwen LLM foundation. It ranks candidate documents by relevance to a query, supports long text and more than 100 languages, and is intended for…
Qwen-Image-2.0-Pro is a full-capability image generation and editing model served through Alibaba Cloud Model Studio. This snapshot was listed in Alibaba Cloud’s release announcement on June 25, 2026;…
Qwen-AgentWorld-35B-A3B is an open-weight language world model designed to simulate environments for AI agents. Given an agent’s action and interaction history, it predicts the next environment…
Unlimited-OCR is Baidu’s open-weight vision-language model for OCR and long-document parsing. The original model was released on June 22, 2026. It is distinct from Unlimited-OCR 2, released later.…
Fun-ASR-Flash is a non-real-time automatic speech recognition model that transcribes audio into text. Alibaba Cloud lists this snapshot under June 17, 2026; the snapshot ID contains the date June 15,…
Qwen3.5-OCR is a Qwen3.5-series optical character recognition model offered through Alibaba Cloud Model Studio. It is designed for document parsing, text localization and key-information extraction,…
Hy-Embodied-0.5-VLA-RoboTwin is a vision-language-action (VLA) checkpoint from Tencent Robotics X and the Tencent Hy Team. It is the supervised fine-tuned RoboTwin variant of Hy-Embodied-0.5-VLA,…
Hy-Embodied-0.5-VLA-UMI is Tencent’s pre-trained vision-language-action (VLA) checkpoint for robotic manipulation. It is a general-purpose starting point for fine-tuning on target robots, rather than…