In partnership with
🖱
Sam Altman just dropped a bomb: OpenAI expects to hit AGI by the end of this year. Yes, 2026. The CEO told TIME magazine they'll have an internal system that qualifies as artificial general intelligence in about four months.
But wait, there's more:
A mysterious "stealth model" called Ox Alpha just got unmasked it's from Chinese AI lab Z.ai
Alibaba launched Qwen3.8-Flash-Next, a coding beast that costs pennies to run
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
Z.ai's stealth model unmasked: Meet Ox Alpha
Sam Altman: We'll hit AGI before 2027
Alibaba drops Qwen3.8-Flash-Next: Coding power, pennies on the dollar
The Mystery "Ox Alpha" Model is Chinese AI GLM-5.3-Flash
The mysterious AI model "Ox Alpha" that everyone loved is actually from China.
Z.ai just claimed it as their new GLM-5.3-Flash model.
This is a huge deal because it matches top US models but costs way less.
The details:
Ox Alpha popped up as a free, anonymous model on OpenRouter and OpenCode
Z.ai just revealed it is a preview of their new multimodal GLM-5.3-Flash model.
The model has 320 billion total parameters but only uses 18 billion at a time to stay fast.
It handles up to 1 million tokens and supports text, images, and video.
Z.ai put the model weights on Hugging Face for anyone to download and change.
Why it matters:
Top-tier AI is getting cheaper fast. Z.ai says this model rivals Claude Opus 4.8 in coding but costs one-tenth the price. By giving away the weights, they are letting developers build powerful tools without relying on expensive, closed-off American companies.
💡 L8R's Take:
I love the stealth launch strategy. It proves the model is good without relying on brand hype. US AI labs need to wake up because China is dropping massive open-source bombs right on their doorstep.
From Our Partner
The CX Architects: Designing the Future of Customer Experience
Running a CX team used to mean following established playbooks. AI changed that overnight, transforming the role before our eyes. In this Career Transformation conversation, host Deepina Kapila talks with three leaders who shaped what came next: Kayla Arp (PrizePicks), Chris Rule (Turno), and Amy Harvey (TrueCommerce), on what changed, the challenges they faced, and where CX is headed. Watch the recording on-demand now.
OpenAI Expects AGI by the End of 2026
Sam Altman just dropped a massive bomb on the tech world.
He told TIME magazine that OpenAI will build an internal AGI system by the end of 2026.
Machines that can think and reason like humans are no longer a distant dream.
The details:
OpenAI Chief Research Officer Mark Chen says the company is already 80% of the way to reaching AGI.
Chief Scientist Jakub Pachocki confirmed OpenAI has hit all of its internal AI research goals.
The company is already using an automated AI research intern named Astra to speed up their work.
This new AGI system will stay internal for now and will not be released to the public right away.
Why it matters:
This puts extreme pressure on rivals like Google and Anthropic to speed up their own work. It also explains why OpenAI is spending billions on custom chips and massive data centers. If they actually hit this goal, the entire tech industry will change overnight.
💡 L8R's Take:
Altman loves to hype things up, but I actually believe the 80% number. You do not spend billions on giant data centers unless you know something huge is coming. The world is not ready for what happens when they flip that switch.
From Our Partners
Your AI budget tripled. See real usage patterns with Harmonic.
AI spend is now a major P&L line item—but most teams can't show what it's producing.
Harmonic Security maps AI activity to use cases and teams, revealing real productivity, shelfware, data risk, and adoption trends across approved and unapproved tools.
Give your board the data behind the return.
Alibaba Drops Qwen3.8-Flash-Next: A Cheap Coding Beast
Alibaba just dropped a massive hint about its next generation of AI.
They released Qwen3.8-Flash-Next, a powerful new model that writes code like a pro but costs pennies to run.
This release acts as a sneak peek into the upcoming Qwen4 family.
The details:'
The model uses a smart design where it holds 125 billion parameters but only turns on 6 billion of them to process a word.
It features a massive 51 billion parameter 'phrase dictionary' that runs on normal computer memory instead of expensive graphics cards.
Alibaba says it beats heavy hitters like DeepSeek-V4-Flash and Anthropic's Claude Opus 4.6 in coding and office tasks.
Developers can process up to 1 million tokens at once if they use a tool called YaRN.
The production version costs a tiny $0.16 per million input tokens and $0.47 per million output tokens.
Why it matters:
Running giant AI models usually burns a hole in your wallet. Alibaba is changing the game by making high-level coding AI insanely cheap. By sharing this design early, they let developers get ready for the massive Qwen4 release. It proves you do not need to use massive computing power to get top-tier results.
💡 L8R's Take:
Alibaba is putting the entire AI industry on notice with this crazy low pricing. If you are building AI software agents right now, you need to test this model immediately. It completely destroys the excuse that good AI is too expensive to run.
🚀 Quick L8R Summary
Altman's AGI timeline: Sam Altman just told TIME he expects OpenAI to hit AGI internally by end of 2026 that's four months away.
Alibaba's Qwen3.8-Flash-Next: Alibaba dropped a new open-weight MoE model with 125B parameters that codes better than most commercial models at a fraction of the cost.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋