- AI Fire
- Posts
- 🤯 Qwen Finally Went Full Omni
Sponsored by
Qwen just went full omni. Alibaba’s new Qwen3.8-Omni-Flash handles text, images, audio, and video with a 1M context window. Audio input costs dropped 98%+.
What's on FIRE 🔥
IN PARTNERSHIP WITH KODEKLOUD
Build. Break. Fix. Learn.
KodeKloud gives you 1,280+ hands-on labs where you provision Kubernetes clusters, write Terraform configs, build CI/CD pipelines, configure Linux systems, containerize apps with Docker, automate with Ansible, and manage Git workflows.
78+ playgrounds let you experiment freely in sandbox AWS environments, Kubernetes clusters, and CI/CD systems without risk.
190+ courses across DevOps, Cloud, and AI pair theory with hands-on labs at every step.
KodeKloud Engineer and 100 Day Challenges provide real-world job scenarios with automated grading that confirms your solutions work.
Stuck? The 55,000+ member Discord community connects you with peers and instructors ready to help.
Every lab runs in a live environment. You deploy, you troubleshoot, you learn. No videos without context. No simulations. The kind of practice that actually builds confidence because you've done real work, not watched someone else do it.
AI INSIGHTS
Qwen3.8-Omni-Flash and its applications in production
Alibaba has released Qwen3.8-Omni-Flash, a natively multimodal model designed to understand text, images, audio, and video inside the same workflow. Main details:
1M-token context window
Accepts text, images, audio, and video
Supports tool calling and built-in web search
Thinking is enabled by default with adjustable reasoning effort
Supports 113 audio languages and dialects
Can process multichannel/spatial audio
Supports context caching and persistent Responses sessions
Maximum text output reaches 131K tokens
Alibaba says performance improved by more than 25% on average across 29 benchmarks compared with Qwen3.5-Omni-Plus. It also claims audio input costs fell by more than 98% per hour!
Qwen3.8-Omni-Flash reportedly beats Gemini 3.8 Flash on several audio and audiovisual tests, but Gemini still leads on some video and agentic benchmarks.
Note: Qwen3.8-Omni-Flash currently has no downloadable open weights announced. Alibaba offers it through its API across regions including Singapore, Virginia, Frankfurt, Tokyo, Hong Kong, and Beijing.
🎁 Today's Trivia - Vote, Learn & Win!
Get a 3-month membership at AI Fire Academy (500+ AI Workflows, AI Tutorials, AI Case Studies) just by answering the poll.
Which statement is currently true? |
PRESENTED BY TIGERDATA
Analytics on Live Data Without Leaving Postgres
When analytics on Postgres slows down, most teams add a second database. Then they manage pipelines, sync lag, and drift forever. TimescaleDB extends Postgres instead. Analytics run on live data, in the database you already have. No pipeline. No migration. No new query language.
HOW TO RUN AI AGENTS FOR FREE
A new model comes out, everyone says it's better, and suddenly you start rebuilding your workflow again.
Fable 5.1. GPT-6 Astra. Tomorrow it could be Codex, Grok, Kimi, or something else.
And every time, you move files, copy prompts, recreate instructions, set up tools again... I hate this. That's why I got interested in this free AI workspace.
It lets you run GPT-6 Astra, Fable 5.1, Codex, Grok, Kimi, and other agents side by side, all using the same files, assets, instructions, skills, and project context.
And everything stays inside the same AI environment. I think this is a much better way to work with AI. In this guide, I'll show you the free AI workspace I'm using, and how I connect multiple models to the same AI OS folder.
TODAY IN AI
AI HIGHLIGHTS
🎁 Muse for Mac is live, plus invite codes. If your friends use your invite code when signing up, both of you get 1 BILLION TOKENS EACH for up to 20 friends! Link here.
🧬 Anthropic is opening Mythos to verified teams, speeding up 30+ biomolecular tools and up to $1M in credits through its program. Any researchers can apply now.
📝 Big Short investor Michael Burry thinks AI danger warnings are being exaggerated by OpenAI & Anthropic ahead of potential IPOs. His breakdown gives 4 reasons.
📢 OpenAI is plugging ChatGPT Ads straight into the tools you already use, with Sponsored Agents, an Ads Manager plugin, HubSpot integration, and a Shopify app.
🏠 Google is rolling out early access to a Google Home MCP server, letting Claude and ChatGPT control Nest cameras, thermostats, and other smart-home devices.
🧠 AI safety debate, again! Microsoft AI’s Mustafa Suleyman is warning about “model welfare,” arguing Claude’s constitution may encourage AI systems to expect rights.
💰 Big AI Fundraising: Crusoe raised $3.9B at a $30.9B valuation to expand huge AI data centers and truck-ready modular “AI factories,” while already holding a $13B, five-year deal with Jane Street.
NEW EMPOWERED AI TOOLS
🤖 Prepathon’26 shows how developers and agencies build REAL, always-on AI agents with OpenClaw, from coding and research to client ops and automation, LIVE and 100% FREE. Join here.
💰 CREEM 2.0 helps AI builders sell software and digital products globally, handling payments, taxes, payouts, affiliates in one platform.
☁️ Bitrise RDE gives AI coding agents cloud Macs that start in seconds, with parallel builds, shared caches, and VS Code or Cursor support.
📱 Modaal lets you ship iPhone and Android apps from one project, with AI generating real code while sharing the same logic.
AI BREAKTHROUGH
Google DeepMind researchers introduced Dream-RSI, a framework that helps AI agents improve how they search for solutions without retraining the underlying model. Main findings:
Dream-RSI learns from the agent’s previous search history instead of repeatedly running expensive new experiments.
It can replay old attempts, simulate alternative strategies, and refine how the agent explores future problems.
The underlying coding model stays unchanged. The improvement comes from the orchestration layer around it.
Researchers tested it across algorithm engineering, mathematical optimization, and GPU kernel engineering.
In some settings, Dream-RSI reportedly reduced the number of costly discovery calls by as much as 162× while maintaining competitive or improved solution quality.
The core idea is simple: instead of starting every difficult search from scratch, the agent turns its past attempts into a training ground for better future exploration. Dream-RSI shows a path toward agents that become more efficient over time.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply