As I write this, 2024 is drawing to a close in an explosion of AI product announcements. Not a day passes without a new model, technique or application being launched or improved. Whether you love it or hate it, AI is here to stay.
That trend of rapid change is expected to not only continue but increase in speed next year as AI models like o3 start to train the next generation of models.
So as we wind down to the holiday season let's cast the dice, take a risky look forward and predict what we may see happening on the AI tree in 2025.
1. Rapid model improvements
All the signs are that the avalanche of new model announcements we've seen this year will slow down somewhat over the next 12 months. That being said, GPT 4.5 or GPT-5 will likely land in some form — maybe as a new merged model with o3.
Meta is already training Llama 4, so that is inevitable and Google has already said new versions of Gemini 2.0 will be released. I'd be surprised if we don't also get Claude 4.0 from Anthropic.
What we don't know about 2025 is whether this models will be versions of the ones we've already seen but with more power and modalities, or if we'll see a shift towards the o1 and o3 models released by OpenAI.
It's highly likely that by the end of next year almost all models will be full voice, vision and text enabled whether they are reasoners or not.
2. The rise of AI agents
More agents, more reasoning and definitely more real world understanding. We're moving away from the world of benchmarks, and arriving at the point where AI has to start earning its keep. That means powering real world tasks, becoming more reliable and delivering more sizzle.
Current research papers hint at various exciting technology breakthroughs, but until they happen we have to assume that we're going to continue to be reliant on traditional transformer chat paradigms.
A hot sector to watch out for next year is that of edge AI computing. The integration of AI into our portable devices and small products is exciting and probable over the coming year.
Google has the most to play for in this regard, with its dominance of the Android smartphone platform. Look out for more complete integration between Android and AI, with Android XR being the start. Apple remains a dark horse.
3. AI applications
The one area where we can expect to see continued and significant progress is that of AI applications. The open source sector continues to build out surprisingly sophisticated and useful tools, and there's no reason that that will slow up over the next year.
In fact there's every chance that the open-source community, congregated around the Llama ecosystem, will start to outpace closed, proprietary applications everywhere except in the enterprise world.
Most of the exciting work in agentic and edge AI applications is coming from this direction, and not just from the West. Suddenly we're seeing important and interesting tools arriving from the Far East.
AI has changed the entire software landscape forever, and this new phenomenon is likely to grow exponentially as the AI tools improve over time.
Final thoughts
Next year will likely see the continued dominance of the current mega AI tech companies. The simple reason is this is an expensive game. Unless you've got a handy billion or more to throw around, it's highly unlikely you're going to be able to enter the market and take on the big guys.
In some ways 2025 is likely to be more of the same, but not. The AI trend is inevitably going to continue its soaring trajectory, but the players may change significantly over the period.
The big surprise may be how prominent the Chinese AI sector becomes over the next months, and also how dominant Google can become if it builds on its recent renaissance in DeepMind products.
Whatever happens, it's not going to be a dull year for anyone, inside or out of the AI industry.
More from Tom's Guide
- Did Apple Intelligence just make Grammarly obsolete?
- 3 Apple Intelligence features I can't live without
- Midjourney vs Flux — 7 prompts to find the best AI image model
