Skip to main content
Back

Tencent unveils its largest open-source text-to-image model

OpenAI's file support in API & parental controls, GitHub Copilot & Copilot Spaces GA, Qwen gets code tools and web sear

September 29, 2025 4 min read

Hey

Welcome to AlphaSignal, the most read source of news by AI engineers and researchers.


Every day, we identify and summarize the top 1% of news, papers, models, and repos, so you’re always up to date.


Here’s today’s roundup:

Summary

Read time: 4 min 36 sec

Top News

▸ Tencent open-sources its largest public text-to-image release

Coresignal

▸ Build AI solutions faster with Coresignal’s multi-source jobs dataset

Signals

▸ OpenAI updates API function calling to support files and images

▸ GitHub makes Copilot coding agent generally available for paid subscribers

▸ Alibaba's Qwen adds code interpreter and web search in chat

▸ OpenAI deploys parental controls to adjust ChatGPT features for teens

▸ GitHub makes Copilot Spaces generally available for all users

Trending Repos

▸ humanlayer: Provide human-in-the-loop workflows for AI decision systems

▸ developer-roadmap: Interactive guides and roadmaps or frontend, backend, DevOps, AI

▸ RAG-Anything: Turn documents into a knowledge graph for better search

Top Read

▸ Google DeepMind finally dropped the Veo 3 paper, the “GPT-3 moment” for video

Top News

Tencent releases open-source HunyuanImage 3.0, matching flagship closed-source text-to-image models in benchmarks

alpha_signal_image_2

Tencent open-sourced HunyuanImage 3.0, its largest text-to-image model to date. With 80B total parameters (13B active per token during inference), it matches the quality of flagship closed-source models while remaining fully open-source.


Key Features

  • Handles long prompts: Understands and follows thousand-word instructions.

  • World knowledge: Uses its multimodal LLM foundation to reason.

  • Text rendering: Accurately generates text inside images.

  • Hybrid training: Combines LLM, diffusion, and multimodal data for integrated capabilities.

How It Works

  • Built on Hunyuan-A13B with Mixture-of-Experts architecture.

  • Uses Transfusion to couple diffusion with LLM training in a unified system.

  • Trained on 5B image-text pairs, video frames, interleaved multimodal data, and 6T text tokens.

  • During inference, only 13B parameters activate per token for efficiency.

Performance

  • Comparable to leading closed-source models in text-to-image benchmarks.

  • Supports complex visual generation like comics, emojis, and detailed illustrations.

  • Strong text rendering inside images, an area where many models fail.

Access & Availability

  • Open-source release on GitHub and Hugging Face

  • Current version supports text-to-image

Try now
alpha_signal_image_3

Build on a complete, multi-source jobs dataset from Coresignal

Stop wasting time on fragmented data. Coresignal’s Jobs Dataset gives you a panoramic view of the labor market with aggregated, deduplicated, and enriched job posting data.


Key Features

  • Multi-source jobs data aggregated from top job boards

  • Duplicate-free, clean, and enriched with company and recruiter details

  • Millions of job listings ready for AI-driven solutions

  • Available via API or as a dataset for integration

  • Power HR tech, sales signals, and market insights with accuracy

Millions of job listings. One dataset. All the details.

Try it now

partner with us

Signals

OpenAI improves function calling in API, let tools return structured file and image outputs

1,439 Likes

GitHub launches Copilot coding agent GA, handling features, bugs, and documentation updates

826 Likes

Alibaba's Qwen integrates code interpreter and web search in chat, enabling instant data analysis

835 Likes

OpenAI rolls out parental controls for ChatGPT, giving parents customizable safeguards and usage limits

1,784 Likes

GitHub's Copilot Spaces is now GA, syncing repositories to keep project knowledge always updated

613 Likes

Repo

humanlayer

4,891 Stars

This repo provides APIs and SDKs to add human oversight into AI agent workflows. It lets you design approval gates, feedback loops, and contextual interventions to make agentic systems safer and production-ready.

developer-roadmap

339,821 Stars

Provides interactive, community-driven roadmaps outlining technologies, tools, and concepts for different developer roles. It helps you navigate learning paths for frontend, backend, DevOps, AI, and more.

RAG-Anything

7,427 Stars

It extends RAG to multiple modalities, including text, images, tables, and equations. It ingests diverse documents like PDFs and Office files, enabling richer query answering and context-aware responses across different data formats.

Top Read

Google finally dropped the Veo 3 paper, the “GPT-3 moment” for video

2,538 Likes

alpha_signal_image_4

Google DeepMind’s Veo 3 paper shows how large-scale video pretraining produces emergent visual reasoning, similar to how LLMs gained language reasoning. The model solves tasks outside its explicit training set, marking a step change in video understanding.


Researchers call it the “GPT-3 moment” for vision.


Key features include:

  • Temporal reasoning across video frames

  • Multi-step inference from raw sequences

  • Generalization to unseen scenarios

  • Transfer to robotics, agents, and simulation

The paper highlights that visual reasoning arises from scale and data diversity, not handcrafted rules. This supports the idea that video models can extend beyond perception into reasoning tasks.


Read the paper or listen to the 15-minute podcast for a quick technical overview.

Read now

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We’re passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.


Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We’re based in the United States. By using our site, you agree to be governed by U.S. laws.

Looking to promote your company, product, service, or event to 250,000+ AI developers?

Work with us