Grok Imagine 2.0 Shines with Layer-Based Editing; OpenAI Secretly Tests GPT-Image-2.5
APPSO · wechat · 2026-08-10
Elon Musk's promoted Grok Imagine 2.0 ranks second globally in text-to-image and image editing on the LMSYS Arena. Its core highlight is interactive precision editing: upon uploading an image, the model automatically segments it into layers (e.g., separating faces, gestures, and backgrounds). Users can directly download subjects with transparent backgrounds or use a magic wand tool to modify specific areas. It also supports automatic stitching and fusion of up to 5 images, widely used for quick meme creation. However, its strict review mechanism has sparked complaints about excessive limitations.
Meanwhile, OpenAI is secretly testing a new image generation model on the Arena under the codename mona-lisa-1. User tests reveal a significant reduction in the "AI feel," producing photos that resemble casual smartphone shots. OpenAI's official verification tool confirmed the generated images contain SynthID watermarks. Speculation suggests this could be the upcoming GPT-Image-2.5 or 2.0 Mini.
More from Models
- Hark Handoff Outperforms GPT-5.5 at 90% Lower Cost — Once_ina_Lifetime · 2026-08-10
- AFP: Chinese AI Drives Price Competition Among US Labs, DeepSeek Tops OpenRouter — 赛博禅心 · 2026-08-10
- Claude Blocks Normal Websites Without Server Requests, Devs Lose Traffic — StereotypicalAussie · 2026-08-10
- Alibaba to Release Open-Weight Qwen3.8-Max, Pressuring US Rivals — emmanuelvivier · 2026-08-10
- Nous Research Releases Open-Source Code Model for Local Execution and Agents — emmanuelvivier · 2026-08-10
- Claude Extended Thinking Bug Silently Burns 674k Tokens — forchat4 · 2026-08-10