The AI updates aren’t slowing down. Literally two days after OpenAI launched a new underlying AI model for ChatGPT called GPT-5.3 Instant, the company has unveiled another, even more massive upgrade: GPT-5.4.

Actually, GPT-5.4 comes in two varieties: GPT-5.4 Thinking and GPT-5.4 Pro, the latter designed for the most complex tasks.

Both will be available in OpenAI’s paid application programming interface (API) and Codex software development application, while GPT-5.4 Thinking will be available to all paid subscribers of ChatGPT (Plus, the $20-per-month plan, and up) and Pro will be reserved for ChatGPT Pro ($200 monthly) and Enterprise plan users.

ChatGPT Free users will also get a taste of GPT-5.4, but only when their queries are auto-routed to the model, according to an OpenAI spokesperson.

The big headlines on this release are efficiency, with OpenAI reporting that GPT-5.4 uses far fewer tokens (47% fewer on some tasks) than its predecessors, and, arguably even more impressively, a new “native” Computer Use mode available through the API and its Codex that lets GPT-5.4 navigate a users’ computer like a human and work across applications.

The company is also releasing a new suite of ChatGPT integrations allowing GPT-5.4 to be plugged directly into users’ Microsoft Excel and Google Sheets spreadsheets and cells, enabling granular analysis and automated task completion that should speed up work across the enterprise, but may make fears of white collar layoffs even more pronounced on the heels of similar offerings from Anthropic’s Claude and its new Cowork application.

OpenAI says GPT-5.4 supports up to 1 million tokens of context in the API and Codex, enabling agents to plan, execute, and verify tasks across long horizons— however, it charges double the cost per 1 million tokens once the input exceeds 272,000 tokens.

Native computer use: a step toward autonomous workflows

The most consequential capability OpenAI highlights is that GPT-5.4 is its first general-purpose model released with native, state-of-the-art computer-use capabilities in Codex and the API, enabling agents to operate computers and carry out multi-step workflows across applications.

OpenAI says the model can both write code to operate computers via libraries like Playwright and issue mouse and keyboard commands in response to screenshots. OpenAI also claims a jump in agentic web browsing.

Benchmark results are presented as evidence that this is not merely a UI wrapper.

On BrowseComp, which measures how well AI agents can persistently browse the web to find hard-to-locate information, OpenAI reports GPT-5.4 improving by 17% absolute over GPT-5.2, and GPT-5.4 Pro reaching 89.3%, described as a new state of the art.

On OSWorld-Verified, which measures desktop navigation using screenshots plus keyboard and mouse actions, OpenAI reports GPT-5.4 at 75.0% success, compared to 47.3% for GPT-5.2, and notes reported human performance at 72.4%.

On WebArena-Verified, GPT-5.4 reaches 67.3% success using both DOM- and screenshot-driven interaction, compared to 65.4% for GPT-5.2. On Online-Mind2Web, OpenAI reports 92.8% success using screenshot-based observations alone.

OpenAI also links computer use to improvements in vision and document handling. On MMMU-Pro, GPT-5.4 reaches 81.2% success without tool use, compared with 79.5% for GPT-5.2, and OpenAI says it achieves that result using a fraction of the “thinking tokens.”

On OmniDocBench, GPT-5.4’s average error is reported at 0.109, improved from 0.140 for GPT-5.2. The post also describes expanded support for high-fidelity image inputs, including an “original” detail level up to 10.24M pixels.

OpenAI positions GPT-5.4 as built for longer, multi-step workflows—work that increasingly looks like an agent keeping state across many actions rather than a chatbot responding once.

Tool search and improved tool orchestration

As tool ecosystems get larger, OpenAI argues that the naive approach—dumping every tool definition into the prompt—creates a tax paid on every request: cost, latency, and context pollution.

GPT-5.4 introduces tool search in the API as a structural fix. Instead of receiving all tool definitions upfront, the model receives a lightweight list of tools plus a search capability, and it retrieves full tool definitions only when they’re actually needed.

OpenAI describes the efficiency win with a concrete comparison: on 250 tasks from Scale’s MCP Atlas benchmark, running with 36 MCP servers enabled, the tool-search configuration reduced total token usage by 47% while achieving the same accuracy as a configuration that exposed all MCP functions directly in context.

That 47% figure is specifically about the tool-search setup in that evaluation—not a blanket claim that GPT-5.4 uses 47% fewer tokens for every kind of task.

Improvements for developers and coding workflows

OpenAI’s coding pitch is that GPT-5.4 combines the coding strengths of GPT-5.3-Codex with stronger tool and computer-use capabilities that matter when tasks aren’t single-shot.

GPT-5.4 matches or outperforms GPT-5.3-Codex on SWE-Bench Pro while being lower latency across reasoning efforts.

Codex also gets workflow-level knobs. OpenAI says /fast mode delivers up to 1.5× faster performance across supported models, including GPT-5.4, describing it as the same model and intelligence “just faster.”

And it describes releasing an experimental Codex skill, “Playwright (Interactive)”, meant to demonstrate how coding and computer use can work in tandem—visually debugging web and Electron apps and testing an app as it’s being built.

OpenAI for Microsoft Excel and Google Sheets

Alongside GPT-5.4, OpenAI is announcing a suite of secure AI products in ChatGPT built for enterprises and financial institutions, powered by GPT-5.4 for advanced financial reasoning and Excel-based modeling.

The centerpiece is ChatGPT for Excel and Google Sheets (beta), which OpenAI describes as ChatGPT embedded directly in spreadsheets to build, analyze, and update complex financial models using the formulas and structures teams already rely on.

The suite also includes new ChatGPT app integrations intended to unify market, company, and internal data into a single workflow, naming FactSet, MSCI, Third Bridge, and Moody’s.

And it introduces reusable “Skills” for recurring finance work such as earnings previews, comparables analysis, DCF analysis, and investment memo drafting.

OpenAI anchors the finance push with an internal benchmark claim: model performance increased from 43.7% with GPT-5 to 88.0% with GPT-5.4 Thinking on an OpenAI internal investment banking benchmark.

Measuring AI performance against professional work

OpenAI leans on benchmarks intended to resemble real office deliverables, not just puzzle-solving. On GDPval, an evaluation spanning “well-specified knowledge work” across 44 occupations, OpenAI reports that GPT-5.4 matches or exceeds industry professionals in 83.0% of comparisons, compared to 71.0% for GPT-5.2.

The company also highlights specific improvements in the kinds of artifacts that tend to expose model weaknesses: structured tables, formulas, narrative coherence, and design quality.

In an internal benchmark of spreadsheet modeling tasks modeled after what a junior investment banking analyst might do, GPT-5.4 reaches a mean score of 87.5%, compared to 68.4% for GPT-5.2.

And on a set of presentation evaluation prompts, OpenAI says human raters preferred GPT-5.4’s presentations 68.0% of the time over GPT-5.2’s, citing stronger aesthetics, greater visual variety, and more effective use of image generation.

Improving reliability and reducing hallucinations

OpenAI describes GPT-5.4 as its most factual model yet and connects that claim to a practical dataset: de-identified prompts where users previously flagged factual errors. On that set, OpenAI reports GPT-5.4’s individual claims are 33% less likely to be false and its full responses are 18% less likely to contain any errors compared to GPT-5.2.

In statements provided to VentureBeat from OpenAI and attributed early GPT-5.4 testers, Daniel Swiecki of Walleye Capital says that on internal finance and Excel evaluations, GPT-5.4 improved accuracy by 30 percentage points, which he links to expanded automation for model updates and scenario analysis.

Brendan Foody, CEO of Mercor, calls GPT-5.4 the best model the company has tried and says it’s now top of Mercor’s APEX-Agents benchmark for professional services work, emphasizing long-horizon deliverables like slide decks, financial models, and legal analysis.

Pricing and availability

In the API, OpenAI says GPT-5.4 Thinking is available as gpt-5.4 and GPT-5.4 Pro as gpt-5.4-pro. Pricing is as follows:

• GPT-5.4: $2.50 / 1M input tokens; $15 / 1M output tokens

• GPT-5.4 Pro: $30 / 1M input tokens; $180 / 1M output tokens

• Batch + Flex: half-rate; Priority processing: 2× rate

This makes GPT-5.4 among the more expensive models to run over API compared to the entire field, as seen in the table below.

Another important note: with GPT-5.4, requests that exceed 272,000 input tokens are billed at 2X the normal rate, reflecting the ability to send prompts larger than earlier models supported.

In Codex, compaction defaults to 272k tokens, and the higher long-context pricing applies only when the input exceeds 272k—meaning developers can keep sending prompts at or under that size without triggering the higher rate, but can opt into larger prompts by raising the compaction limit, with only those larger requests billed differently.

An OpenAI spokesperson said that in the API the maximum output is 128,000 tokens, the same as previous models.

Finally, on why GPT-5.4 is priced higher at baseline, the spokesperson attributed it to three factors: higher capability on complex tasks (including coding, computer use, deep research, advanced document generation, and tool use), major research improvements from OpenAI’s roadmap, and more efficient reasoning that uses fewer reasoning tokens for comparable tasks—adding that OpenAI believes GPT-5.4 remains below comparable frontier models on pricing even with the increase.

The broader shift

Across the release and the follow-up clarifications, GPT-5.4 is positioned as a model meant to move beyond “answer generation” and into sustained professional workflows—ones that require tool orchestration, computer interaction, long context, and outputs that look like the artifacts people actually use at work.

OpenAI’s emphasis on token efficiency, tool search, native computer use, and reduced user-flagged factual errors all point in the same direction: making agentic systems more viable in production by lowering the cost of retries—whether that retry is a human re-prompting, an agent calling another tool, or a workflow re-running because the first pass didn’t stick.

Nominal on Thursday announced a fresh $80 million Series B extension round at a $1 billion valuation, led by Founders Fund. This followed the company’s $75 million Series B round led by Sequoia in September.

The company offers software that helps hardware engineers test their designs and began as a picks-and-shovels type of startup for the defense industry. The three-and-a-half-year-old, L.A.-based company says that in the last 10 months, it has landed four of the five largest defense contractors as customers.

CEO and co-founder Cameron McCord (pictured) is a former U.S. Navy submarine officer and an alum of defense tech company Anduril, which is also one of Nominal’s marquee customers. So perhaps it’s no surprise that Anduril co-founder and executive chairman Trae Stephens, who is also a Founders Fund partner, led this as a preemptive deal, McCord told Bloomberg and confirmed to TechCrunch. Sequoia, General Catalyst, Lux Capital, Red Glass, and Lightspeed also participated.

Next up, Nominal plans to expand beyond defense tech and into industrial sectors like automotive, robotics, and other industries. And it’s made a good start. The company tells TechCrunch that some of its other customers include Pratt Miller Motorsports (the Corvette Racing Team) and nuclear energy company Antares.

Data from Hogan Assessments explores how falsehoods about women in positions of power can result in missed opportunities for all.

Ahead of International Women’s Day (8 March), Hogan Assessments, a HR consulting platform, released globally sourced data on the personality traits typically linked with effective leadership in the workplace. 

What was discovered is that men and women in executive roles often demonstrate similar personality traits linked to leadership effectiveness, challenging long-standing assumptions about who is best suited to lead. Despite this, Allison Howell, the CEO of Hogan Assessments, noted that women are frequently underrepresented in leadership roles because of assumptions that they lack the ‘necessary’ traits. 

Howell told SiliconRepublic.com: “This assumption stems from outdated models of leadership that equate effectiveness with visible confidence, dominance or self-promotion. These traits have historically been overvalued, shaping perceptions of potential. 

“However, the data shows that men and women demonstrate similar capabilities on the traits that actually predict leadership success, such as judgement, integrity and emotional regulation.”

For Howell, to make an assumption is to give oxygen to unfounded and even harmful ideas. 

“The most effective way to challenge them is through evidence,” she said. “Decades of personality and performance data consistently show no meaningful gender differences in leadership effectiveness. 

“Organisations need to align their definitions of potential with what the science actually supports. That means evaluating people based on how they actually lead teams and deliver results, not on how confidently they present in a meeting.”

Hogan’s research suggests that for many women, the real challenge is less about readiness and more about access. The report stated that “early promotion and development decisions are often shaped by informal judgements that quietly influence who is encouraged, supported and advanced over time, long before leadership roles are formally in sight”.

With that in mind, Howell is of the opinion that to undermine assumptions and challenge falsehoods, a culture of shared responsibility must be created from within the core of an organisation. She explained, companies have to critically evaluate their own systems and criteria, while leaders at every level should question biases in decision-making. 

She believes there should also be room for key mentorship and sponsorship as it often plays a critical role in increasing access and visibility. She said: “Seeing others navigate similar paths can help demystify leadership journeys. At the same time, support networks are most effective when paired with fair and transparent systems.”

Bridging the ambition gap

Hogan’s research also explored the ‘ambition gap’, which Howell explained is the incorrect assumption that women in the working environment sometimes have less ambition than their male co-workers. 

She explained that what is often labelled as a lack of ambition is in fact a reasonable response to structural workplace conditions. She said: “Our data shows that women don’t lack ambition, women show comparable drive and competitiveness. 

“However, ambition may appear differently between men and women because of the need to adapt to the environment. For example, if an organisation rewards men for competing with colleagues but punishes women for the same behaviour, women are more likely to adapt to the environment.”

Noting that structural barriers often shape leadership pipelines long before the C-suite stage, Howell said they can emerge early through informal judgements via hiring, promotion and development opportunities. “These moments influence who gains visibility, support and stretch assignments. Over time, small disparities compound, shaping who ultimately reaches senior leadership.”

For Howell the research makes it clear that the barriers are “real, not imagined”, but knowing that “can be freeing”, as it gives women and indeed those aiming to empower women in the workplace the knowledge needed to break down barriers blocking access to leadership positions.  

She said: “Progress requires both reflection and action. The evidence is clear, leadership effectiveness isn’t defined by gender, but by how well we align potential with performance. 

“Rethinking outdated assumptions is essential to building more inclusive and effective leadership pipelines. The question isn’t whether or not women can lead. It’s whether organisations are willing to recognise leadership when it doesn’t look like what they are used to seeing.”

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

The Jolla Phone makes its grand premiere at MWC 2026 with a calm air of unmistakable Finnish flare, a mobile device that reinvigorates the spirit of freedom in phone manufacturing. However, this is no average smartphone. Jolla has painstakingly designed a cellphone centered on Sailfish OS, its own in-house Linux-based system, and matched it with some strong mid-range components that feel really practical for 2026.

At the center of this is Sailfish OS, specifically version 5, an operating system that can function without the need for a Google account, avoids background data collection that feeds advertising networks, and avoids all of the hidden analytics that assist push adverts. You can’t blame Jolla for repeatedly highlighting this point, because the system only sends data back to headquarters if the user requests it. A tactile switch on the side allows you to turn off the microphone, cameras, Bluetooth, and other sensors instantaneously, which is a wonderful touch. You won’t find that amount of control on many phones in people’s pockets.

Sale

Google Pixel 10a – Obsidian – 128 GB with $100 Amazon Gift Card

• Order the new Google Pixel 10a today and get an Amazon Gift Card; valid 2/18/2026 until 3/11/2026 at 11:59pm PT, while supplies last and subject to…
• Return of Pixel 10a without gift card results in charge; offer cannot be combined, is non-transferable and not valid for cash or cash equivalent
• If a qualifying item in your order is returned, you’ll be reimbursed for what you return, minus the value of the gift card

The hardware selections keep everything in check; there’s no over-the-top flagship stuff here. The phone has a 6.36 Full HD+ AMOLED display, wonderfully covered by Gorilla Glass, with 390 pixels per inch and a stunning 20:9 aspect ratio. MediaTek’s Dimensity 7100 5G chip powers the show, delivering solid performance for daily tasks and connectivity without overheating drama. Configurations offer 8 GB or 12 GB of RAM from SK hynix alongside 256 GB of storage, which expands up to 2 TB via microSD card. Dual nano-SIM slots sit alongside that card slot in the tray.

The cameras use Sony sensors, including a 50-megapixel main snapper and a 13-megapixel ultra-wide on the back, as well as a nice wide-angle front lens (listed as approximately 32 megapixels if you look closely). The results should be adequate for regular usage, however Jolla’s emphasis is on dependability rather than sophisticated camera technology. The 5,450 mAh to 5,500 mAh battery is completely user-replaceable, a deliberate tribute to longevity that some manufacturers have all but abandoned. Swappable back covers are available in Snow White, Kaamos Black, and an attractive orange. These colors mirror the original Jolla Phone’s module concept, and the pogo pins for future accessories designed by the community are a wonderful tribute to the spirit of the original “The Other Half” concept.

Connectivity is extensive, including 5G compatibility, Wi-Fi 6, Bluetooth 5.4, NFC, and positioning via GPS, Galileo, GLONASS, and BeiDou. The dimensions are around 158 x 74 x 9 mm, making this phone a good fit in the pocket, with flat corners and minimal bezels.

App access bridges the gap that once seemed impossible to bridge, as Jolla’s AppSupport technology now runs a wide range of Android apps with ease, including banking tools and messaging services. Of course, compatibility isn’t flawless, and certain programs behave strangely, but the configuration prevents you from accessing full Google services unless you add MicroG as an optional extra. This allows people to stick to their preferred routines without having to deal with the normal tracking mess.

Pre-orders, however, tell a quite different story. This thing debuted in December 2025, and while starting with a modest goal, it quickly grew to over 10,000 in commitments and a tidy €5 million in sales by the end of February 2026. A new batch of 1,000 units arrived at MWC, still available for €649 each, albeit you’d need to put down a €99 deposit to get in on the action, with September deliveries slated for the EU, the UK, Norway, and Switzerland. The first exports are expected at the end of June 2026. Jolla is offering a few signals that they have bigger plans in the future – think ecosystem growth and some extremely fascinating AI capabilities that are tied in with all their work on keeping things private.
[Source]

Last week I got to go see some of what LG is bringing to the marj in 2026 with its TVs and sound systems. I’ll talk about the TVs in a later article but what caught my eye (or ears) was its Sound Suite audio system.

You may know that I have a degree of angst about TV sound. More often than not, it’s not very good and isn’t likely to get better in 2026. The best solution is usually to rely on a soundbar or surround sound system and LG’s Sound Suite is, interestingly, both these things.

Flexible and modular

That’s because the Sound Suite system is modular, which I’ll harp on about again because it makes more sense than relying on a single soundbar. As long as you have space.

You can mix and match satellite speakers, subwoofer, soundbar – even your LG TV – and have them work together in concert. Like a sports team, they understand each other’s roles and can intelligently adapt.

There’s quite a number of combinations on offer. You could use the TV and soundbar, the soundbar and subwoofer; the TV and four satellite speakers, or all joined together as one harmonious whole. What if I wanted to take two satellite speakers and play in stereo? I can do that, and then when I’m done, reassign them as part of the group of speakers again.

There’s flexibility so you can start small and build up, adding speakers when you need (and if you need them). I can place these speakers anywhere I’d like in a room – the speakers can figure out where they are in relation to one another, and expand and contract the soundstage based on their proximity.

Each speaker has a microphone, which allows it to measure the distance from the soundbar (or TV), understand where it is relative to others, and measure the room acoustics. Do I need to worry about reflections in a room affecting the sound? According to Dolby, not really.

If I move a speaker, recalibrate (which takes twenty seconds), I can also see the new position of the speaker on the LG ThinQ app.

And it fixes an issue I’ve long held about Dolby Atmos, which is that in its current incarnation, it’s too front-heavy to be immersive. The idea of selling me a soundbar, and trying to make me believe that it can fling sounds all across the room (including behind me), is a non-starter.

But the ability to add satellite speakers, which some soundbars such as the Marshall Heston 120 or KEF XIO don’t even support, opens it up for a true surround sound experience.

Dolby FlexConnect finally arrives

And this is possible because of Dolby Atmos FlexConnect, which was introduced a few years ago and then put in its best impression of The Fugitive by being elusive and not very easy to track down.

TCL and Hisense announced FlexConnect support when it was first announced and then, if memory serves, neither really mentioned it much afterwards.

So this is, two years later, the first time I’ve actually heard a FlexConnect system in the wild. Instead of this rigidity that traditional 5.1 or 7.1 surround system would enforce on me, I can now place speakers so they match the layout of my room. There’s potential for some good audio Feng Shui here.

And another pet peeve is how brands market surround systems where you’ve got a sofa dead centre in the middle of the room surrounded by plenty of open space. I know not of any living room that is that big or has that much space.

FlexConnect understands that no room is perfect and compensates. It can blend sound across physical speakers to make it sound like it’s coming from a specific area, so there’s no need to position speakers perfectly. What if I rearrange furniture in a room – will that affect the sound? Again, that’s not expected to be an issue.

It’s a clever system but the proof is in the pudding and judging from the time I spent listening to it, the sound depends on the speakers you have involved. I’m still not convinced about the sound from LG’s own TVs, so when partnered with the soundbar, it does sound thin and lacking weight and warmth. It sounded a little too brash.

But with the H7 Soundbar, W7 subwoofer, or the M5 and M7 satellite speakers engaged, and then you have something closer to a ‘proper’ sounding surround system. Poweful, energetic, expansive and immersive. It might be a bit too crisp for some, but this seems to be the tuning that LG is, and has gone for, for quite some time.

The price is… alot

The price though, is paramount. The H7 Soundbar is £899, the M7 satellite speakers are £399 a pop. For well over £1500, you only get a soundbar and two surround speakers. This is premium and compare that to the Samsung HW-Q990H which will have the soundbar, rear speakers and a subwoofer for probably around £1500, and suddenly the LG Sound Suite doesn’t look as good in terms of value.

But if there’s anything that would convince, it’s the ease of having Dolby Atmos FlexConnect. I’ll be looking forward to see how good it sounds, once I’ve got the system set-up.