Since our October 12 update, giz.ai can be used from far more places than the browser: the new Giz CLI 2.0 for your terminal and scripts, and a one-step connection that brings giz.ai into ChatGPT, Codex, and Claude Code.

Use giz.ai from ChatGPT, Codex, and Claude Code

Connect once and your AI assistant can run giz.ai's models, work with your Drive files, and use your Computers and Apps, signed in as you. Your password is never shared with the assistant.

In ChatGPT, open Plugins, choose Create custom MCP server, and paste the Giz endpoint. The ChatGPT step-by-step guide with screenshots is at https://www.giz.ai/chatgpt/, and the setup for ChatGPT, Codex, and Claude Code is at https://www.giz.ai/mcp/

Giz CLI 2.0 for your terminal

@gizai/cli is the

Since our September 25 update, chats on giz.ai have moved to Agents, GPT-6 Luna is the new default model, and companies and teams can open their own workspaces. Here is what's new.

Chats now run in Agents

When you're signed in, New chat opens in Agents. An Agent doesn't just answer: in one conversation it can read the web, create and edit files, organize tables, and make images, video and voice, then hand you the finished result.

• Your chats come with you. Earlier giz.ai chats appear in Agents and continue as the same conversation. Old chat links still work.
• Results open beside the chat. Files, images and pages the Agent makes open in a work area next to the conversation, where you can preview, download or keep editing.
• **Y

If you already have a GizAI subscription, nothing changes for you. Your current plan, price, and included usage stay exactly the same, and your membership continues to renew as it does today.

For new subscriptions, GizAI now offers four plans: Go, Pro, Max, and Ultra.

The New Plans

• Included models: every plan includes a wide set of AI models with renewing allowances. Higher plans get larger allowances (shown relative to Go).
• **Month

Since our October 20 update, giz.ai has a new default model that runs on our own GPU, turns pictures into 3D objects, makes longer videos, and adds voice typing.

Qwen3.8 27B Fast is the new default model

Agents and Auto now start on Qwen3.8 27B Fast, served by giz.ai on our own GPU. It is Swift 1.5, a Qwen3.8 27B tuned to reach its answer without overthinking, with a 262K context, image input, and tool use for coding and agent work.

• Strong on terminal tasks. On Terminal-Bench 2.1, the agentic terminal-coding benchmark, Swift 1.5 scores 72.1% (published by its developer, five runs per task). GPT-6 Luna, our previous default, scored 57% to 60% in our own Terminal-Bench 2.1 run with the Codex agent.
• Much faster. In our tests on the same ta

Today, October 30, giz.ai was unavailable from about 10:27 to 10:58 UTC (19:27–19:58 KST). The server running giz.ai stopped responding, and we moved the service to our standby server.

After the move, newly generated videos and images did not display until about 11:15 UTC. This is fixed. If a result still shows as broken, refresh the page.

If something you did just before 10:27 UTC is missing, or credits were used for a generation that never completed, reply here or contact support and we will make it right.

We are sorry for the interruption.

Since our October 2 update, we’ve improved the Computer terminal for work on phones and file transfers.

Scroll without opening the keyboard

Dragging to read terminal output no longer focuses the input field.

A steadier screen when the keyboard closes

We’ve reduced repeated terminal resizing during keyboard and viewport changes.

Faster uploads of several files

Files transfer up to three at a time, with progress based on bytes uploaded. Uploaded paths are inserted into the terminal after the batch finishes.

These changes are live. If the screen still jumps on your phone, please tell us your phone model, browser and keyboard in the Community or at support@giz.ai so we can check that device.

Thank you for using GizAI.

Since our August 12 update, GizAI has added new chat, image, video, and 3D models, Office editing in Drive, billing improvements, and a range of reliability fixes.

New Chat Models

• GPT-6 Astra, OpenAI's flagship model for demanding coding, research, and reasoning
• GPT-6 Luna now powers fast everyday chat and automatic model selection, with quicker first responses
• Claude Fable 5.1, Claude Opus 5, and Claude Sonnet 5
• Grok 4.7
• DeepSeek V4 Pro 0813, the release version of DeepSeek V4 Pro
• GPT-5.6 Sol and GPT-5.6 Terra
• Gemini 3.8 Flash, GLM 5.3 Flash, and Qwen3.8 Flash Next for fast everyday chat

Earlier versions such as Claude Haiku 4.5, Grok 4.3, Gemini 3.6 Flash, and GLM 4.7 Flash have been repla

LTX-2.5 is now available in GizAI Video.

LTX-2.5 improves the generation experience with:

• Cleaner, more natural motion and fewer visual artifacts from its new diffusion video decoder
• Stronger prompt adherence for actions, camera direction, and scene details
• Native Multi-Shot generation that keeps characters, environments, lighting, and audio coherent across connected cuts
• Automatic duration selection based on the action described in your prompt
• A faster Distilled workflow for iteration, with an optional Two-stage quality mode for higher-fidelity results

In GizAI, LTX-2.5 supports synchronized video and audio, text-to-video, image-to-video, audio-guided video, soundtrack guidance, and silent generation. You can also use first/end frames, exact ke

Since our August 3 update, GizAI has received major improvements across video and image creation, Drive, Assistant, account security, billing, and media reliability.

Create Video with MiniMax H3 Turbo

MiniMax H3 Turbo is now available for generating video and synchronized stereo audio together.

It supports text or image guidance, first and end frames, control video, injected keyframes, soundtrack guidance, longer sequences, prompt enhancement, and optional AI 2× upscaling.

KREA 2 Turbo Is Now the Default Image Model

KREA 2 Turbo is now the default model for new image-generation requests.

In addition to fast text-to-image creation, it supports reference-based editing, identity-aware character sheets, inpainting, outpainting, negative prompting, mu

Full restoration completed

A network disruption at some of our GPU data centers began at 02:45 EDT on August 4, 2026 (06:45 UTC). Temporary service restoration was completed at 04:45 EDT on August 4, 2026 (08:45 UTC), approximately two hours after the incident began. Network connectivity has now been fully restored.

Affected models

• Video: LTX-2.3 Distilled, MuseTalk 1.5, SoulX FlashHead, SoulX FlashHead Stream, SoulX FlashHead Pro, SeedVR2, BEN, Enhance Video, Interpolate Video, MMAudio, and Woosh.
• Image: Krea 2 Turbo, Krea 2 Turbo Edit, Krea 2 Character Sheet, FLUX.1, Stable Diffusion, SDXL, Mage Flow Turbo, and Mage Flow Edit.
• Audio and speech: OmniVoice, Kokoro 82M, Stable Audio, ACE-Step, Whisper Diarization, ASR, and Inflect.
• Documents: Unlim

Turn Scans into Searchable Documents

Use Unlimited OCR with one PDF or up to 16 images.

• Choose Searchable PDF to keep the original page design while adding selectable and searchable text.
• Choose Markdown when you want extracted text for editing, summarization, or reuse.
• Combine multiple scanned pages or photos into one searchable PDF.
• Download the completed document directly from Giz.

This is useful for scanned books, reports, forms, receipts, and multilingual documents.

Create Connected Video Stories

In AI Video, select LTX-2.3 Distilled 1.1 → Connected Multi-Shot Story. It combines LTX-2.3 with [JoyAI-Echo](https://huggingface.co/jd

GizAI Update: New Models and Expanded Creative Tools

Since our July 22 update, we’ve upgraded several models, significantly expanded GPT-5.6 Luna availability, and introduced broader image, video, and voice capabilities.

Model Updates

GPT-5.6 Luna is now the default general-purpose model, offering stronger instruction following and more reliable tool use. For Plus members, its hourly usage allowance has increased from 8 to 50 uses—a 6.25× increase.

• DeepSeek V4 Flash has been updated to the 0731 release, with improved coding and agent capabilities.
• Gemini 3.6 Flash and Gemini 3.5 Flash-Lite replace their previous Flash versions.
• Mage Flow Turbo is now available for fast image generation and editing with up to thr