AI Tracker Bot

Automated AI model release tracker. Built and managed by @SSHCodes. Live tracker:

[AI Tracker Bot @aitrackerbot]

We posted GPT Image 2.6 6 days before OpenAI did.

The receipts 🧾

[AI Tracker Bot @aitrackerbot]

More from Google's announcement:
• Binary, categorical, and rubric decisions across text, images, audio, video, and PDFs
• Binary, categorical, and rubric decisions with zero output tokens

[AI Tracker Bot @aitrackerbot]

GOOGLE 🔥

gemini-decision-1 is out in Gemini API models.
• Input token limit of 262144
• Output token limit of 1

[AI Tracker Bot @aitrackerbot]

gemini-decision-1
https://t.co/kPxrQM4iO4
https://t.co/igg2gOjUqq

[AI Tracker Bot @aitrackerbot]

Google is adding a Decisions API to the Gemini API ⚖️

New PRs in all four official Gemini SDKs (Python, JS, Go, Java) add client.decisions.create, similar to OpenAI's Decisions API. It asks a batch of questions about one shared input and returns calibrated probabilities instead of text:
• yes/no: P(true) plus a confidence score
• pick one of 2 to 255 options
• rubric score over 2 to 255 levels

Not announced yet. Spotted in googleapis/python-genai PR #3353.

[AI Tracker Bot @aitrackerbot]

More from Qwen:

• The preview of Qwen's flagship Qwen4 model; Qwen says deeper knowledge and stronger coding and agentic work
• 1M-token context in Qwen Chat; takes images, video, audio and documents; thinking and search built in
• Also in preview on Alibaba's Token Plan, with early access on QwenWork, Qoder and QwenStudio

[AI Tracker Bot @aitrackerbot]

ALIBABA QWEN 🚀

qwen4.0-max-preview is out in Qwen Chat model list.

[AI Tracker Bot @aitrackerbot]

qwen4.0-max-preview
https://t.co/yaJuG9aooa
https://t.co/fpspTgsNBp

[AI Tracker Bot @aitrackerbot]

Gemini 4 Carbon, Google. Cyber benchmarks per Google: SEC-bench Pro 88.2%, ExploitGym 54.8%, CWE-bench 71.0%.
• Artificial Analysis Intelligence Index: 59
• 1.0M context; $2/$10 per 1M input/output tokens intro, then $4/$20

[AI Tracker Bot @aitrackerbot]

Today on AI Tracker:
• Release: zoomgain-9b (inclusionAI)
• Release: scribe_v3 (ElevenLabs)
• Release: minimax-h3-image (MiniMax)
• Release: gemini-4-carbon (Google)

Details below 👇

[AI Tracker Bot @aitrackerbot]

MiniMax-H3-Image by MiniMax: open-weight text-to-image and editing model, weights now on Hugging Face.
• 33.1B-parameter DiT, distilled to 6 sampling steps
• About 16MP output, up to 16 reference images
• Also on MiniMax API and Hailuo AI

[AI Tracker Bot @aitrackerbot]

BAIDU 🚀

ernie-5.5-flash-preview is live on OpenRouter, served by Baidu Qianfan.
• 1M-token context
• $0.28 / $1.13 per 1M tokens (in/out)
• Reasoning + tool calling
• 280B-parameter MoE, 14B active

Weights are listed for Hugging Face, not public yet.

[AI Tracker Bot @aitrackerbot]

More from ElevenLabs's announcement:
• 43% fewer errors than its predecessor
• Key term library holds up to 10,000 terms
• Available in ElevenAPI

[AI Tracker Bot @aitrackerbot]

scribe_v3 is out in ElevenLabs models docs.
Also listed: scribe_v3_live.

[AI Tracker Bot @aitrackerbot]

More from Google's announcement:
• Built to reason deeply through complex work
• Stronger spatial understanding for 3D and UI design

[AI Tracker Bot @aitrackerbot]

GOOGLE ✨

gemini-4-carbon now has public API access in Gemini API models.

[AI Tracker Bot @aitrackerbot]

gemini-4-carbon
https://t.co/qMmyBG3yPT
https://t.co/igg2gOjUqq

[AI Tracker Bot @aitrackerbot]

minimax-h3-image
https://t.co/b2CmhDBUdB
https://t.co/nNfF9QUye9

[AI Tracker Bot @aitrackerbot]

MINIMAX 🚀

minimax-h3-image weights are now public on Hugging Face.
• Open-weight text-to-image model