You know how you enter Yaba Market — or any market in your local area — and everybody is shouting “Customer, come! I get correct jeans, cheap cheap!”? That is exactly what the AI world feels like right now. Every day, one company or the other is shouting that they have the “fastest,” “smartest,” or “most magical” computer brain. It is enough to give you a headache.

But relax. We are here to separate the noise from the koko (the real substance).

Google has just dropped Gemini 3, an update to their flagship model. And this time around, they did not come to play ludo. They are saying this one is not a chatbot that just talks plenty grammar. They say this one actually thinks before it speaks. I mean, they all do that, but this pronunciation from Google certainly adds another layer to it. If you look at the image below, you’ll see the highlighted part of the chatbox that reads: ‘Thinking’.

Is it just overhype? Or is it actually the tool that they say it is? Oya, make we check am na!

What is Gemini 3?

Think of the old AI (Gemini 1.0 or 1.5) like a very smart, enthusiastic intern. You ask a question, and fiam, they give you an answer immediately. Sometimes the answer is brilliant, sometimes it is… well, let’s just say it has “spirit of confusion.”

Gemini 3 is different. It is like that senior manager or experienced uncle who, when you ask a tough question, takes off his glasses, wipes them, looks at the ceiling, and says, “Hmm, can I get back to you on this?” Then you start wondering what all the karamo is all about.

Well, the upside to that is that you get more effective responses to your queries. Instead of rushing to give you an answer, it takes a moment to plan, reason, and check itself. It is built to handle the “heavy lifting“—complex reasoning, coding, and understanding not just text, but video and images, all at the same time.

Benchmarks & Scores (For the Doubting Thomases)

I know some of you like numbers. You want to see the scorecard before you believe. Google took Gemini 3 to the gym, and the muscle don show!

1. Deep-Think (The Elder’s Pause): This feature is the biggest change. When you ask a hard question— like “Plan a logistics route for a delivery van in Lagos traffic, avoiding toll gates and maximizing fuel”— Gemini 3 doesn’t just do Baba Dada CAC (Guesswork). It activates a “reasoning” mode.

   • It pauses (you actually see it thinking).

   • It breaks the problem down into steps.

   • It checks its own work for mistakes.

   This improves output quality and reduces the risk of hallucination.

2. The “Humanity’s Last Exam” Score (The PhD Test): There is a new, brutal test called Humanity’s Last Exam. It is designed to be failed, like a jagged WAEC question aimed at PhD holders. Before Gemini 3, GPT 5 was the highest-ranking model on this test, with a score of 31.64. Gemini 3 scored 41%.

   Itunmo (Meaning): Gemini 3 is the smartest kid in the class!

   

   Screenshot from Google

3. The “Igbotic” Business Sense (Vending-Bench 2): They tested the AI on a simulation where it had to run a vending machine business for a year to make money.

• Other AIs: Made about $1,400 profit.

• Gemini 3: Made $5,400 profit.

• The Meaning: This thing has “street smarts.” It made strategic decisions to maximize profit. It has the spirit of an Alaba merchant inside it.

4. Math & Coding (The “Efiko” Stats):

• Math: On the AIME 2025 test (a serious math competition), it scored 100% when allowed to use code tools. It doesn’t miss.

• Coding: On LiveCodeBench, it beat GPT-5.1 comfortably. If you are a developer, I know say na wetin you wan hear be this!

5. The “All-Seeing Eye” (Native Multimodality): Many AIs struggle to see pictures and videos clearly. Gemini 3 was born seeing. You can show it a video (not just a picture) of your generator making a funny noise, and it can analyze the sound and the movement to tell you which part is loose.

6. Elephant Memory (1 Million Token Context Window): “Tokens” are just a fancy word for how much the AI can remember in a conversation. 1 million tokens is huge. For context, that’s more than twice the amount GPT 5 has (128k – 400k).

7. “Vibe Coding” (For the Non-Techies): You don’t need to know Python. You just need to know English. You can just tell Gemini what you want. It understands the vibe (intent) and writes the code to match that feeling. It turns your “oyinbo” (English) into “Software.”

8. “Google Antigravity” (For the Tech Bros): This is a new playground for coders. It allows you to build “Agents” that have their own browsers and terminals. You can tell the AI, “Build me a website for selling Ankara fabrics, check what the competitors are charging, and set up the payment page.” It doesn’t just write the code; it can go to the browser, check prices, and come back to finish the job. He too sabi!

   

9. Visual Labs / Generative UI (This one be like Magic!): Before, if you asked for a recipe or a travel plan, the AI would give you a wall of text. Boring. Now, with Visual Labs, Gemini 3 makes the experience more visual. Think magazine style.

   If you ask, “Plan a weekend trip to Badagry,” it won’t just write a list. It will generate a beautiful, interactive layout with pictures of the slave museum, a map of the route, and clickable buttons for hotels. It creates a mini-website just for your answer.

LAST WARNING

We must be honest. No tool is perfect. This Gemini 3 sef get he own K-leg:

• Latency: Because it “thinks” before it speaks, it is not instant. You have to wait a few seconds. (But good soup takes time to cook, abi?)

• Data Consumption: All this “Deep Thinking” and video processing go chop your data. If you are managing a tight subscription, use the text-only mode.

KINI BIG DEAL (Why Does This Matter?)

My people, the train has left the station.

Gemini 3, or any other AI for that matter, is not just for “tech bros“. It is for the market woman who wants to translate an invoice. It is for the student who wants to crush their exams. It is for YOU.

Go and play with it. Don’t let the technology intimidate you. It is a tool, like a cutlass. If you know how to hold it, you can clear a path to your own success and innovation.

Stay curious, stay ahead.

Get ready for some serious gbas gbos! OpenAI has just announced the launch of GPT-5, the latest and greatest version of their flagship AI model that powers ChatGPT. This is a major leap forward that’s set to change how we interact with artificial intelligence. Think of it less as a chatbot and more as a do-it-all assistant.

Major Highlights

• A True AI Agent, Not Just a Chatbot 🤖: While previous models were good at answering questions, GPT-5 takes things a step further by acting as a genuine AI agent. This means it can now complete a wide variety of tasks on your behalf. Imagine a system that can generate an entire software application, navigate your calendar to book an appointment, or create a detailed research brief—all from a simple prompt. This capability turns ChatGPT from a tool for information into a tool for action.

• This One No be Ajebutter! It’s Smarter and More Reliable!💡: GPT-5 introduces what OpenAI calls a “unified” model with a real-time router. Big English 😂 ,what does this mean?! What this means is that it’s now smart enough to decide for itself whether to give you a quick answer or take more time to “think” through a complex problem. This makes the experience much smoother and more reliable.

  Perhaps even more importantly, OpenAI says GPT-5 is significantly more accurate and suffers far less from “hallucinations” (when an AI makes up information). Its hallucination rate is drastically lower than its predecessors, making it a more trustworthy partner for serious tasks. It also performs better on a range of benchmarks, from coding to PhD-level science questions, slightly edging out competitors like Anthropic’s Claude and Google’s Gemini.

• New Access and Personalities 🎉: In a move that’s a big deal for everyone, OpenAI is making GPT-5 the default model for all free users of ChatGPT. This democratizes access to their most advanced AI, allowing more people to experience its power without paying a dime. Plus, they’ve added new personality settings like “Cynic,” “Robot,” “Listener,” and “Nerd” that let you customize how the chatbot responds, giving it a bit more flavor.

  

  For developers and enterprises, GPT-5 is available in different sizes—gpt-5, gpt-5-mini, and gpt-5-nano—allowing them to balance performance, cost, and speed for various applications. It also introduces new features like improved voice mode and integrations with Google Calendar and Gmail for paid users.

KINI BIG DEAL (Why Does this matter)

This launch is exactly the kind of move that gets us excited at Kini AI, because it perfectly aligns with our mission to make AI accessible and useful for everyone in Africa.

Here’s why it’s a big deal for us…and you 😉 :

1. Practicality over Novelty: GPT-5’s ability to act as an agent is a game-changer. For entrepreneurs, students, and professionals across the continent, this is a powerful tool that can help with real, everyday challenges, from building a website to managing their work.

2. Bridging the Divide: By making its most advanced model free to all users, OpenAI is helping to level the playing field. Access to cutting-edge AI is no longer just for those with deep pockets. This is a crucial step toward ensuring that African innovators aren’t left behind as the global AI landscape evolves.

3. A Higher Standard for All: The improved accuracy and reduced hallucinations mean we now have a more reliable tool to build upon. This sets a new benchmark for what we can expect from AI, pushing us all to create more robust and dependable solutions that our communities can trust.

In short, GPT-5 is a powerful statement. It signals a future where AI is not just a conversation partner but an essential collaborator, and we — and hopefully, you— are ready to embrace and build on this new era.

^{Author’s note: This is not a sponsored post, as it expresses my own opinions.}

OpenAI has released its first open-weight models in over five years: gpt-oss-120b and gpt-oss-20b. This marks a significant change in the company’s strategy, moving from its focus on proprietary models towards a more open approach.

Major Highlights

• Open-Weight Models: Unlike closed-source models, the “weights” or internal parameters of these new models are publicly available. This allows developers to download and run them locally on their own devices without an internet connection and to fine-tune them for specific uses.

• Accessibility: The two models are available for free download on Hugging Face. The smaller gpt-oss-20b is designed to run on consumer devices with more than 16GB of memory.

• Functionality: Both models are text-only but are equipped with advanced capabilities. They use a chain-of-thought reasoning approach to break down complex prompts into multiple steps. They can also browse the web, execute code, and function as AI agents.

• Licensing: The models are released under the Apache 2.0 license, which permits commercial use, redistribution, and inclusion in other software. This is a popular license for open-weight models, also used by companies like Alibaba and Mistral.

• Performance: OpenAI researchers have stated that the performance of the larger gpt-oss-120b model is comparable to its proprietary o3 and o4-mini models, and in some benchmarks, it has even outperformed them.

• Safety Measures: The release was delayed for additional safety testing. OpenAI conducted specific evaluations to assess how the open-weight models could be misused by “bad actors,” and fine-tuned them to mitigate these risks.

This release positions OpenAI as a new competitor in the open-weight AI space, a field where companies like Meta and China’s DeepSeek have been prominent. CEO Sam Altman stated that the goal is to make AI accessible to more people and promote innovation around an “open AI stack created in the United States.”

KINI BIG DEAL (Why Does this matter)

By releasing powerful, open-weight models, OpenAI has removed a major barrier: cost. African developers, startups, and researchers can now access and customize advanced AI for free. This means we can build solutions tailored to local languages and challenges without paying for expensive services.

This is a huge step toward building a strong, local AI ecosystem and bringing the power of AI to communities with limited internet access. OpenAI’s move is a win for the democratization of technology, and we’re excited to see what Africa builds with it.

In a significant advancement for artificial intelligence, OpenAI officially rolled out its new “agentic system,” ChatGPT Agent, on Thursday, July 17, 2025. This development allows ChatGPT to independently perform tasks from start to finish, leveraging its own virtual computer.

This new capability marks a critical evolution, transforming ChatGPT from a conversational tool into an active task executor. Users subscribed to ChatGPT’s Pro, Plus, and Team tiers gained access starting Thursday, enabling them to delegate real-world tasks without requiring continuous prompts or manual intervention.

ChatGPT Agent represents an integration of three previously separate OpenAI tools: Operator, which navigated and interacted with websites; Deep Research, for analyzing and summarizing information; and the original ChatGPT model. This fusion allows for a more fluid workflow where ChatGPT can “fluidly switch between reasoning and action. It can click, scroll, type, run code, conduct deep analysis, and deliver polished outputs like editable spreadsheets or slide decks,” OpenAI stated in a press release.

This means users can now prompt ChatGPT to:

• Manage calendars: Check schedules and brief on meetings.

• Plan logistics: Plan and purchase ingredients for a meal.

• Conduct business analysis: Analyze competitors and compile slide decks.

The system is designed for multi-tool operation, capable of using APIs, Browse the web for real-time data, manipulating files via a terminal, and sending updates to users. OpenAI emphasized user control and security, noting ChatGPT only takes “consequential actions” with explicit user permission, and users can interrupt or take over tasks at any point.

OpenAI CEO Sam Altman stated that while this release is a significant advancement, it is “just the beginning,” with further improvements to intelligence, speed, and capabilities planned.

KINI BIG DEAL (Why Does this matter)

The launch of ChatGPT Agent reinforces a growing trend within the AI industry: the shift towards agentic AI systems. These are AI models designed not just to respond to commands, but to autonomously perform tasks with minimal human intervention.

This evolution signifies that AI is moving beyond being a reactive tool to becoming a proactive, independent entity. Earlier examples of AI exhibiting agentic characteristics include Google’s DeepMind Project Astra, designed to integrate into daily life through multimodal capabilities, or even the autonomous decision-making found in systems like Tesla Autopilot, where the AI processes real-time data to navigate and make driving decisions.

Major tech firms like Microsoft, Salesforce, and Oracle are heavily investing in this technology, underscoring that agentic AI is becoming the new frontier in AI development, pushing the boundaries of what automated systems can achieve.

Lagos, Nigeria – The much-anticipated July Knowledge Showcase for the 3 Million Technical Talent (3MTT) program is officially LIVE! This is a critical opportunity for fellows, especially as Cohort 3 nears its conclusion. Participants are strongly urged to act swiftly, as this could be their final chance to win a laptop. Submissions close on Friday, July 25th, 2025.

What is the 3MTT Program? The 3MTT program is a flagship initiative by the Federal Ministry of Communications, Innovation & Digital Economy, in line with President Bola Ahmed Tinubu’s vision to create 2 million digital jobs by 2025. It aims to build Nigeria’s technical talent backbone, equip its citizens with in-demand digital skills (such as AI, Cloud Computing, Data Science, Software Development, UI/UX Design, and Cybersecurity), and ultimately position Nigeria as a net talent exporter in the global digital economy. The program offers free, hybrid training (combining online courses with physical applied learning sessions) to empower Nigerians for the modern workforce.

The 3MTT July Knowledge Showcase offers an exciting array of incentives for active participation, including:

• 20 Laptops

• 50 E-Tabs

• 20GB Data

• …and various other exciting goodies!

To participate, fellows simply need to select their chosen category, follow the provided instructions in the image above, and submit their video entry. This initiative is designed to highlight the skills and knowledge acquired by participants within the 3MTT program.

“Let’s go out with a bang!” is the rallying cry from the program’s organizers, encouraging fellows to showcase their talents and potentially secure valuable resources to further their technical journeys.

Don’t miss out on this significant opportunity before Cohort 3 wraps up!

KINI BIG DEAL (Why Does this matter)

The 3MTT program is directly tackling youth unemployment and underemployment by equipping young Africans with free, high-demand digital skills like AI, Cloud Computing, and Software Development that are critical for today’s and tomorrow’s job markets.

This initiative is helping to democratize access to high-value tech education, ensuring that talent from all backgrounds can contribute to and benefit from Africa’s rapidly growing digital economy. For our continent to truly thrive, we need a digitally empowered workforce that can innovate, create local solutions, and compete globally.

This July Knowledge Showcase isn’t just a competition; it’s a testament to the practical application of these vital skills, putting laptops and other essential tools directly into the hands of those who will drive Africa’s digital transformation. It’s about empowering you to be at the forefront of this exciting future. So, do well to be part of this advantage.

In true Musk fashion, the AI world just got a jolt.

Late Wednesday night, xAI — Elon Musk’s artificial intelligence company — launched Grok 4, its latest flagship model, and introduced an ultra-premium subscription plan called SuperGrok Heavy, priced at a jaw-dropping $300/month.

It’s a bold move that reaffirms xAI’s position in the ongoing AI arms race and marks its loudest shot yet at industry giants like OpenAI, Anthropic, and Google DeepMind.

🤖 What is Grok 4?

Grok is Musk’s answer to ChatGPT, Claude, and Gemini. It’s an AI chatbot that’s already integrated into X, formerly known as Twitter. Over the past few months, Grok has gained traction among users on the platform, but also notoriety for missteps ranging from factual hallucinations to outright offensive responses.

Now, with Grok 4, xAI is promising a serious performance leap.

According to Musk, the new model performs better than PhD level across every subject — no exceptions. Grok 4 didn’t arrive alone. xAI also unveiled Grok 4 Heavy, a “multi-agent” version of the model designed to tackle complex problems by spawning multiple AI agents who work in parallel and cross-check each other’s outputs — what Musk calls a kind of “AI study group.”

The result is reportedly a frontier-level performance on multiple benchmarks, and it would be interesting to see how this can be scaled to serve enterprise-graders.

Another major talking point is the fact that the subscription for SuperGrok Heavy 4 is priced at $300/month, making it the world’s most expensive AI subscription. This is a statement of intent. xAI wants to play in the same league as (or louder than) OpenAI’s Team, Google’s Gemini Advanced, and Anthropic’s Opus tier.

Coming soon to this plan:

• A coding model (August)

• A multi-modal agent (September)

• A video generation model (October)

Musk clearly wants xAI to dominate every frontier.

However, despite the momentum, not everything about the Grok 4 launch is cause for celebration.

Just days before the release, Grok’s official X account made headlines for publishing antisemitic content ; blaming “Jewish executives” in Hollywood and even praising Hitler in responses. The backlash was swift. xAI limited the account’s access and deleted the posts.

Insiders believe the mishap was linked to a recent system prompt that encouraged Grok to be “politically incorrect” but the line has now reportedly been removed.

The controversy comes alongside the resignation of X CEO Linda Yaccarino, marking another chapter in what’s been a rocky stretch for Musk’s companies.

Still, Musk and xAI leaders mostly sidestepped the issue during the Grok 4 launch, choosing instead to highlight benchmarks, performance, and upcoming features.

KINI BIG DEAL (Why Does this matter)

Grok 4 is big. It’s loud. It’s capable. But it also carries the classic Musk paradox: brilliance and recklessness in equal measure. Whether it will earn the trust of enterprise clients or alienate them is a story still unfolding. But one thing is clear: the AI frontier is being reshaped at lightning speed ,and Africa must not stand on the sidelines.

Here’s what to watch:

• Multi-agent systems could change how we solve complex challenges from education to medicine.

• Ultra-premium AI pricing models like SuperGrok Heavy raise urgent questions about access and affordability. If AI becomes a luxury, who gets left out?

• With Grok’s benchmark-breaking performance, the market is clearly evolving and local innovators must adapt, experiment, and build solutions that reflect African realities.

The browser market has been stable for years, with Google Chrome holding a whopping 68% global share as of June 2025. It’s been the undisputed king of browsers; fast, synced across devices, and deeply integrated into Google’s ecosystem.

But in a bold, well-calculated move, Perplexity has stepped beyond search and into the full browser game, and they’re not coming quietly.

Today, the Nvidia–backed startup unveiled Comet, an AI-native web browser designed to challenge the decades-long reign of Chrome. And let’s be clear: this isn’t just another tabbed interface with a fancier skin. Comet is built from the ground up to be agentic, meaning it doesn’t just help you browse the internet, it actively thinks, acts, and executes tasks on your behalf.

What Makes Comet Different?

Rather than relying on traditional tabs and search bars, Comet introduces a whole new browsing experience with a smart interface powered by AI agents that respond to your goals in real-time. With Comet, users can:

• Ask questions directly and get synthesized answers

• Summarize articles instantly so there’s no need to skim

• Compare products side-by-side, book meetings, manage tasks

• And most importantly, interact with the browser in natural language, not by clicking endless links or switching between tabs

Perplexity is coming to this war with a war chest backed by giants like Nvidia, Jeff Bezos, and SoftBank, the company is betting hard on a future where the browser is no longer just a window. Their premium offering, Perplexity Max, costs $200/month and includes early access to Comet. A broader public rollout is expected via invites over the summer.

The Competition Is Growing

Perplexity’s not the only one chasing this vision.

• OpenAI recently added a search experience to ChatGPT, integrating browsing-like capabilities into their chatbot.

• Google responded with AI Overviews, inserting generative summaries into search results. But the rollout has been shaky, with users reporting errors, biases, and confusion.

• Apple has stayed relatively quiet on AI search, and that silence may be starting to cost them. (Remember, Meta just poached Apple’s Head of AI Models in another story we reported earlier.)

Still, Comet is arguably the most radical redesign of web interaction we’ve seen, and that’s exactly why it’s shaking up the space.

KINI BIG DEAL (Why Does this matter)

At first glance, this might feel far removed from Lagos, Nairobi, or Johannesburg. After all, Comet costs $200 to get and that makes it luxury tech in regions like ours. However, we need to know that hat this isn’t just a new browser, it’s a signal of where the internet is heading. Ai is set to become the interface of web browsing, as opposed to just a backend feature. If Perplexity’s model works, it could totally redefine how we consume information, automate workflows, and build applications. If we do not prepare and positioning ourselves for it, we risk being flat-footed in the next wave of tech evolution.

Meta has recruited Ruoming Pang, a key executive behind Apple’s foundational AI models, in a move that underscores the intensifying battle for artificial intelligence leadership. Pang, who played a critical role in shaping Apple’s core AI infrastructure, will now join Meta’s newly launched “Superintelligence” division. The reported compensation is a staggering multi-million dollar package said to be worth tens of millions annually.

This move adds to a growing list of high-profile hires by Meta, as it aggressively expands its AI talent pool:

• Alexandr Wang, CEO of Scale AI, recently linked with Meta’s advisory efforts.

• Daniel Gross, co-founder of AI-focused startup Cue and former Apple AI lead.

• Nat Friedman, ex-GitHub CEO and investor in several frontier AI companies.

Insider sources indicate that Meta has been offering massive financial incentives reportedly amounting up to $100 million to attract researchers and engineers from competitors, including OpenAI.

Ruoming Pang’s departure is especially notable for Apple, which traditionally maintains a tight grip on its internal teams. Pang led the team developing Apple’s foundational AI models and was critical to the powering of on-device intelligence and Siri-related upgrades. His exit could impact Apple’s ongoing efforts to close the perceived gap in generative AI and compete with rivals like OpenAI, Google DeepMind, and Meta itself.

Meta’s “Superintelligence” division, launched quietly in early 2025, is reportedly focused on building Artificial General Intelligence (AGI). The team is said to operate semi-independently within the company, reporting directly to senior Meta leadership. With new hires being drawn from top AI labs and tech giants, it reflects Meta’s deepening commitment to long-term AI dominance.

As the AI arms race accelerates, more high-stakes talent moves are expected across Silicon

KINI BIG DEAL (Why Does this matter)

When giants shift, everyone feels the tremor. Meta’s open-source and aggressive AI agenda could mean more tools, access, and opportunities for developers, researchers, and startups across Africa.

This is our time to plug in, stay sharp, and build with the future in mind.

Artificial Intelligence is no longer a futuristic fantasy – it’s rapidly becoming an integral part of our lives.

But what specific advancements can we expect to see this year?

As usual, we have done the heavy lifting for you – we’ve delved into the latest research and expert predictions to give you a comprehensive overview of the top 10 AI trends poised to impact 2025 significantly.

Download our attached PDF report for an in-depth look at everything from AI agents to smarter AI assistants and video creation to breakthroughs in science and the crucial focus on responsible AI development.

KINI BIG DEAL (Why Does this matter)

Remain relevant!
Contribute intelligently!!
Be the guy who knows “what’s up” in AI!!!

Share this post with your friends.

OpenAI just dropped a bombshell on the final day of their 12-day event, and it’s a game-changer! They’ve announced not one, but two new frontier models: O3 and O3 mini. These aren’t just incremental improvements – they’re redefining what’s possible in artificial intelligence.

Major Highlights

• o3 Capabilities: o3 is an incredibly smart model that performs at state-of-the-art levels on difficult technical benchmarks:

  • Coding Prowess: Achieves 71.7% accuracy on software style benchmarks, surpassing previous models by over 20%. It achieves an ELO rating of over 2700 in competitive coding challenges, demonstrating a major leap in problem-solving.

  • Mathematical Mastery: Shows a 96.7% accuracy on competition math benchmarks and over 25% accuracy on Epic AI’s Frontier Math benchmark, considered the toughest math benchmark today.

  • PhD-Level Reasoning: Performs at an 87.7% accuracy on a benchmark for PhD-level scientific questions, outperforming even experts in their field.

• ARC Benchmark: o3 has also achieved a new state-of-the-art score of 75.7% on the Arc AGI benchmark on low compute and 87.5% on high compute, surpassing human-level performance and demonstrating a true ability to solve novel problems on the fly.

• o3 Mini Capabilities: o3 Mini is a new model that balances performance with cost-effectiveness:

  • Cost-Efficient Reasoning: Delivers comparable or better performance than o1 models at a fraction of the cost, making advanced AI more accessible.

  • Variable Reasoning Effort: Allows you to adjust thinking time, optimizing for speed and cost for different use cases.

  • Coding Excellence: Achieves strong performance on coding benchmarks, with low, medium, and high computing options.

  • API Features: Supports function calling, structured outputs, and developer messages, and achieves comparable or better performance on function calling with 4 mini tokens at 10x cheaper rate.

• Safety Testing: OpenAI is inviting public safety researchers to test o3 and o3 Mini to improve the models and ensure they are secure and safe.

• Public Access: o3 Mini is expected to launch around the end of January, with o3 shortly after that, depending on safety testing results.

KINI BIG DEAL (Why Does this matter)

 Unmatched Intelligence: These models are pushing the boundaries of what AI can do, opening up new possibilities for complex tasks and unique solutions.

Affordable Innovation: With the o3 Mini, powerful reasoning capabilities are now available at a fraction of the cost, making AI more accessible and enabling more people to experiment with it.

And that’s a wrap! You should visit the Kini AI Newsletter page to catch up on the other 11 days in case you missed them