openai Archives - Page 2 of 14

OpenAI and Microsoft drop exclusivity

OpenAI’s models can now be distributed through AWS and Google Cloud. (Picture: Microsoft)

As Microsoft made big early bets on OpenAI, they gained exclusivity to their models and a revenue share from OpenAI.

Now, with a $50 billion deal from Amazon waiting in the wings, the companies are amending the deal so the models can be made available «across any cloud provider.»

That’s big news for enterprise adoption and competition with Anthropic, as their models can now be made widely available on Google Cloud and AWS.

As part of the new deal, Microsoft is still OpenAI’s «primary cloud partner,» and will get dibs on new models. The difference is the non-exclusive license through 2032, and that they will no longer pay a revenue share to OpenAI. OpenAI, however, will be paying Microsoft through 2030, with an undisclosed total revenue cap.

— With this, builders will have even more choice to pick the right model for the right job, Amazon CEO Andy Jassy wrote, according to Reuters.

Read more: OpenAI’s announcement, Reuters, and Ars Technica

OpenAI is building a «fundamentally changing» agent-first phone, Kuo says

This is how Ming-Chi Kuo imagines a new agent-first interface. (Picture: Ming-Chi Kuo)

Supply chain analyst Ming-Chi Kuo, mostly known for breaking early Apple news, has published an article on x.com claiming that OpenAI is already working with MediaTek and Qualcomm to develop a phone with an agentic interface.

— Smartphones will remain the largest-scale device category for the foreseeable future, he writes, and shipments of high-end phones are around 300-400 million units a year — a mass market OpenAI would be keen on reaching.

The phone itself should be an agent-first experience, «fundamentally changing how people think of smartphones,» Kuo says.

This means moving away from apps to agents, from icons to tasks, and then doing away with the well-worn grid interface in favor of an agent-powered stream layout.

Along with the chip giants on board, systems integrator Luxshare is slated as their main manufacturing partner, aiming for a full system spec late this year or early 2027. Mass production is «expected» in 2028, meaning the project is moving fast.

Read the full story on x.com.

OpenAI moving to a faster, «iterative» release schedule for new models

We can expect more models at a faster pace from OpenAI going forward. (Picture: Adobe)

GPT-5.5 came out just six weeks after GPT-5.4, so naturally, people are wondering if this is part of a new trend of quicker releases:

— Yes, we expect quite rapid continued progress, says OpenAI Chief Scientist Jakub Pachocki on a call related to today’s release, according to Tae Kim on Substack.

— We see pretty significant improvements in the short term, extremely significant improvements in the medium term, he adds.

Sam Altman has also weighed in, saying that «We believe in iterative deployment; although GPT-5.5 is already a smart model, we expect rapid improvements.»

Altman goes on to say that this is also part of a safety strategy, arguing that iterative releases make it easier for the world to prepare for advances in AI.

Read more: Key Context by Tae Kim, Altman’s tweet.

OpenAI releases GPT-5.5, a step up in intelligence with the same latency

From chatbot to «research partner,» ChatGPT-5.5 steps up with complex reasoning. (Picture: Adobe)

Some advanced models lose speed when upgraded, but not so for OpenAI’s latest. GPT-5.5 handles more demanding tasks at a faster clip — and «excels» at agentic coding, computer use, knowledge work and scientific research.

It also matches the GPT-5.4 on token latency, «while performing at a much higher level of intelligence,» OpenAI says.

Codex is much improved by the new model, with better code debugging, document and spreadsheet creation, use of software and moving across toolsets.

It’s also strong on benchmarks, using significantly fewer tokens to complete the same Codex tasks as GPT-5.4.

On computer use and scientific research, it should get higher quality results with fewer tokens or retries, and requires less guidance, OpenAI says.

The model is more expensive than GPT-5.4, clocking in at $5 per million input tokens and $30 per million output tokens — almost twice the cost.

Read more: OpenAI’s announcement, writeups at CNBC, Axios and The Verge.

OpenAI makes it super easy to create and deploy agents in ChatGPT

You just need a description of the task to get going with ChatGPT’s new agents. (Picture: OpenAI)

Dubbing them «workspace agents,» OpenAI is rolling out Codex-powered agents to Business, Enterprise, Edu, and Teachers tiers.

In a bid to popularize agents beyond enthusiasts and IT departments, these agents are for everyone, and are sharebale between teams.

They are also very easy to create; simply add a description of the job you want done, and ChatGPT walks you through the process of turning it into an agent.

Continue reading “OpenAI makes it super easy to create and deploy agents in ChatGPT”

OpenAI brings web search, thinking to ChatGPT Images 2.0

A stoat on a goat on a boat in a moat, approaching photorealism. (Picture: generated)

The new image model for ChatGPT does not only offer higher fidelity, but more precision in how it renders pictures. According to OpenAI, it’s a sea change:

— If we think of Dall-e as cave drawings, and Images 1.0 as ancient art, then Images 2.0 is the Renaissance, OpenAI claims, according to Gizmodo.

It is a little bit faster to generate, and offers better instruction following, more accurate object placement — and reasoning with web search.

That means it can render multiple images from a single prompt, and «double-check» its outputs, as well as offering the latest information in the picture.

As for the benchmarks, GPT Image 2.0 has entered LMArena’s leaderboards at #1 for Text-to-Image with a whopping 242 point lead, and is 125 points ahead on Single-Image-Edit and up 90 on Multi-Image Edit.

The model is available today across all of ChatGPT’s tiers. For the thinking mode, you’ll need a paid Plus, Pro or Business subscription.

Read more: OpenAI’s announcement, Launch thread, Gizmodo and The Verge.

OpenAI introduces «Chronicle,» a tool constantly screenshotting your desktop

The new feature is tailored to high-output work environments, or it would be a privacy disaster. (Picture: Adobe)

The new feature is an agent observing your screen all the time you work, storing screenshots as «memories» to better help with context for your Codex tasks.

— Over time, it helps Codex learn how you work: the tools you use, the projects you return to, and the workflows you rely on, OpenAI croons on x.com.

The point is to learn even more detail about you, from how you prefer your code to the tools and apps you use to perform. This can then later be recalled by Codex.

Notwithstanding the privacy concerns from Windows Recall, which also uses AI to take and store screenshots of your desktop, OpenAI is warning that the screenshots are even stored unencrypted on your computer.

They also warn that it eats up rate limits quickly, is very prone to prompt injection attacks and is only available on the $200 Pro subscription, as a research preview on macOS. Once enabled, it can be paused at any time in a menu item.

OpenAI launches GPT-Rosalind, specifically made for life sciences, biology

OpenAI wants to aid in research and discovery of new drugs, but hallucinations linger. (Picture: Adobe)

Aiming to help discovery and create therapies from vast databases and cutting edge research, access to the new model will be tightly restricted.

For obvious reasons, biohacking can be a serious issue even for general AI implementations, but when it comes to building a model strictly for biology, only a select few researchers will get access.

The Rosalind model is based on the latest internal research from OpenAI, and outperforms GPT-5.4, sometimes massively, on chemistry, biochemistry, genetics and experimental design — the datasets it was trained on.

Continue reading “OpenAI launches GPT-Rosalind, specifically made for life sciences, biology”

OpenAI releases revamped Codex app with computer use for macOS

Codex is getting one step closer to a super app. (Picture: OpenAI)

Getting one step closer to their super app, OpenAI’s latest Codex app can operate every app on your computer by «seeing, clicking and typing» with its own cursor — in the background (so you don’t have to wait for it to finish).

The app can also now generate images, remember preferences, and learn from previous workflows. It even comes with its own in-app browser — so you can check your web work instantly.

It can also open PDFs, spreadsheets, slides and docs natively, and gets a new summary pane to track agents, sources and «artifacts,» in addition to alpha support for SSH connections and multiple terminal tabs.

As a «preview,» it should be able to reuse older threads for context and instructions, and schedule its own work over days and weeks.

The app is available here. Computer use only works on the Mac version.

Read more: OpenAI’s announcement, The Verge, TechCrunch.

OpenAI announces ChatGPT-5.4-Cyber, a permissive vulnerability researcher

Only the most trusted cybersecurity pros will get access to the advanced model. (Picture: OpenAI)

The new Cyber model has fewer restraints than other available bots to let cybersecurity professionals game out and test for vulnerabilities.

These kinds of tasks would normally get refusals for security reasons, but with Cyber access, developers can go as far as reverse engineering entire apps to poke for bugs.

The model is based on ChatGPT-5.4, but OpenAI says they are expanding the entire Cyber program now «in preparation for increasingly capable models over the next few months […] whose capabilities will rapidly exceed even the best purpose-built models of today.»

The release comes hot on the heels of Anthropic’s Mythos model and «Project Glasswing,» said to be so advanced they won’t release the full model.

To get access to GPT-5.4-Cyber, you have to first verify that you are a cybersecurity professional with OpenAI, and even then you might get «limited» access based on a tier system.

Read more: OpenAI’s announcement, Reuters and Axios.

OpenAI memo touts «Spud,» says Anthropic is «single-product company»

OpenAI has great faith in their upcoming «Spud» model, and thinks Anthropic blew it on compute. (Picture: generated)

Chief Revenue Officer Denise Dresser revealed OpenAI’s priorities going forward in a recent memo seen by The Verge — and it’s all about enterprise and their upcoming model.

— Better model performance lifts the rest of the stack. «Spud» will make all of our key products significantly better, writes Dresser, without giving any timeline for release.

She also touts their compute advantage, saying it will show up in higher token limits, in training stronger models, and as a better fit for multi-year, multi-functional customer needs.

The enterprise markets that OpenAI has been focusing on of late are maturing, she writes, and says customers aren’t necessarily looking for the latest and greatest, but the best fit, the best workflows, and day-to-day operations, where they hope «Spud» will deliver.

She then has some choice words on Anthropic; saying it was a strategic misstep «not to acquire enough compute,» that their focus on coding makes them «a single-product company in a platform war,» and that «their story is built on fear, restriction.»

Read more: The Verge has the memo, Axios, CNBC focuses on the Amazon business.

OpenAI expects ad revenue of $2.5 billion this year, and $100B by 2030

If OpenAI continues on its growth trajectory, they project massive earnings from ads. (Picture: Adobe)

Counting on continuing the pace of user growth, OpenAI is banking on ad revenues well above the more cautious estimates of $1 billion, Axios reports, citing sources privy to investor presentations.

The growth to $100 billion revenue will be possible if they hit a target of 2.7 billion weekly users by 2030 — and would put it in contention with some of the world’s largest advertising platforms, notes Reuters.

Behemoths like Google and Meta has advertising revenues of $295 and $196 billion annually as of today, and OpenAI’s planned trajectory would make AI chatbots one of the pillars of online marketing.

Previously, OpenAI shared a projected $100 million revenue from ads after just six weeks of operation, and said they were building a self-serve advertising platform.

Read more: Axios and Reuters.

Sam Altman reflects on AGI and policy after Molotov attack on his home

Altman is calling for a cooling of the debate. (Picture: Shutterstock)

Comparing Artificial General Intelligence to The One Ring from Tolkien’s legendary books, Altman says he understands the heated debate about AI and who controls it:

— A lot of the criticism of our industry comes from sincere concern about the incredibly high stakes of this technology, he writes, and — Once you see AGI you can’t unsee it.

The OpenAI CEO strikes a humble tone in reflecting on the attack, which saw a Molotov cocktail thrown at his house, which luckily bounced away. San Francisco police later apprehended a man in his twenties.

Continue reading “Sam Altman reflects on AGI and policy after Molotov attack on his home”

OpenAI announces $100 ChatGPT Pro plan, with 5x limits by popular demand

Codex growth is off the charts for OpenAI, which is now releasing a custom-made subscription. (Picture: OpenAI)

ChatGPT is on par with Claude subscriptions, now that it has an intermediate Pro subscription tailor-made for Codex use.

This also closes the huge gap between the Plus tier at $20/month and the top Pro tier at $200/month.

Sam Altman says the new tier is by very popular demand. A spokesperson tells TechCrunch that 3 million people are using Codex every week, and it’s growing by more than 70% per month.

To celebrate the launch of the new tier, OpenAI is increasing usage limits for the $100 plan to 10x for Codex, while the $200 plan remains at 20x.

OpenAI will «for sure» reserve stock for retail investors in IPO, CFO says

In an interview with CNBC, Chief Financial Officer Sarah Friar stays mum on specifics of going public, but says their recent funding round showed «really strong demand» from retail investors.

She said other companies, like Block and SpaceX have reserved IPO stock for retail investors and OpenAI will «for sure» do the same thing.

— Everybody wants to own part of a rocket company — I hope everyone wants to own part of ChatGPT. It helps when you’re a consumer brand, Friar told the channel.

She won’t comment on a concrete timeline for going public, but says it’s «good hygiene» for the company to «look and feel and act like a public company.»

An OpenAI IPO could come as soon as the second half of 2026, Reuters reports, and could value the company at up to $1 trillion.

Read more: CNBC, Reuters.