Some of our IFA 2026 highlights

The IFA 2026 technology trade event is currently underway in Berlin, Germany and the Tom's Guide team is there in force to check out the latest announcements. So far we've seen stair-climbing robots, an exoskeleton that folds down into a backpack and a Motorola phone packed with Swarovski crystals.

Of course, there's also a ton of new laptops and we've been getting more hands-on time with Nvidia's impressive RTX Spark. Here are some of our highlights.

• I used Dell's latest MacBook Neo rival — and I'm not complaining about having another one
• I went hands-on with Lenovo’s RTX Spark-powered Yoga 9n 2-in-1 laptop — and it’s a creative powerhouse
• I tested Acer's wild Project DualPlay Mini — and it's a gaming handheld and laptop in one gadget

GTA 6 controllers

Grand Theft Auto VI Limited Edition DualSense Controllers | PS5 - YouTube

Watch On

There's a new PS5 controller coming, and it's fully decked out in GTA 6 styling. It's not as in-your-face as I expected, but it does feature the main color scheme we've seen across all the Grand Theft Auto 6 content released so far.

I think they look pretty awesome, and I will definitely consider picking one up. But what about you? Sound off in the poll below and let us know whether the GTA 6 PS5 controller tickles your fancy or is a pass.

Anthropic: Fable 5.1 is 'world's most advanced' model for knowledge work

Anthropic has launched its Claude Fable 5.1 model with a claim that it's the "world's most advanced" for coding and knowledge work. The company's own benchmarking tests say the new model improves on Fable 5 in four key areas: agentic scientific research, agentic terminal coding and multidisciplinary reasoning, and agentic coding.

Crucially, it's also meant to be cheaper to run. In a blog post announcing the new model (and its restricted access version, Mythos 5.1), Anthropic wrote: "Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads, wherever usage is billed by token. This is because we’re reducing our pricing on cache reads (where the model reads inputs that have already been processed and stored). For highly agentic work, the savings will often be much larger—up to approximately 45%."