See the history
behind AI models

GPT-6 Cyber launched under the distinct API ID `gpt-6-cyber` for authorized vulnerability research, with API and Codex access for approved Daybreak Red customers.

Google made Gemini Omni 1.2 Flash (`gemini-omni-1.2-flash`) generally available in the Gemini API and Google Cloud on 2026-11-05, with video generation up to 15 seconds and multi-turn conversational editing.

ERNIE 5.5 Flash Base released model weights.

OpenAI launched ChatGPT Images 2.6 and released the `gpt-image-2.6-flare` and `gpt-image-2.6-sunburst` image generation and editing models for developers.

Tencent Hunyuan’s official Research feed dates the Hy4 formal release to 2026-11-03. The originator previewed it for long-horizon productivity and first-time visual understanding; GA parameter count, weights, and license are not disclosed.

Google announced Gemini 4 Carbon on 2026-10-28, beginning controlled access for a subset of cyber defenders and planning to expand to trusted partners; developer, business, and consumer access is forthcoming.

Genie 4 entered research preview with trusted testers for interactive world generation and native video and audio output; broader user and developer access is still forthcoming.

Microsoft opened the MAI Image 2.7 image-editing preview on Arena, Artificial Analysis, and MAI Playground; text-to-image is still forthcoming.

OpenAI launched GPT-6.1 Terra (API ID `gpt-6.1-terra`) for visual reasoning, computer use, and everyday professional work, with availability through the API, ChatGPT Work, and Codex.

Moonshot launched Kimi K3.1 Pro as the first model in the K3.1 series, available on Kimi products, mobile apps, Kimi Work, Kimi Code, and the Kimi API; full weights are promised by Nov 4.

Tencent released Hy-Nomas code and Apache-2.0 weights on 2026-10-23 for multi-turn video-world generation controlled by text and camera paths.

MiniMax launched M3.1 Flash through the MiniMax API, M Plan, and MiniMax Code for agentic work, coding, and computer use.