Google upgrades Gemini 2.5 Pro preview ahead of GA
Google just rolled out an upgraded preview of Gemini 2.5 Pro before its general availability in a couple of weeks. The update gained 24 Elo points on LMArena to hit 1470, improves formatting and creative writing, and introduces thinking budgets in Google AI Studio to let developers manage latency and cost.
Google is giving its smartest model one final push right before general availability.
Google rolled out an upgraded preview of Gemini 2.5 Pro. Building on what was shown at I/O in May, this iteration will transition into the fully stable, enterprise-scale GA version in just a couple of weeks.
Benchmark jumps and thinking budgets
The update pushed Gemini 2.5 Pro even further ahead on public leaderboards. It jumped 24 Elo points on LMArena to 1470, keeping the top spot, while climbing 35 points on WebDevArena to 1443. Google also noted top-tier performance on demanding reasoning and coding benchmarks like Aider Polyglot, GPQA, and Humanity’s Last Exam.
Beyond raw intelligence, Google addressed previous user complaints by polishing the model’s formatting and creative style. Developers can test it now in Google AI Studio and Vertex AI, where Google also introduced adjustable thinking budgets to rein in cost and latency. The update is hitting the Gemini consumer app today as well.
Why it matters
In our view, handing developers a knob for thinking budgets is the smartest operational upgrade here. Frontier reasoning models are notoriously expensive and slow when left unchecked; giving teams precise control over how long the model ponders makes real-world enterprise adoption far more viable.
With GA slated just two weeks away, Google is aggressively locking in its lead on coding and multi-step reasoning. It appears determined to prove that Gemini 2.5 Pro is not just an arena benchmark champion, but a production-ready engine ready for enterprise budgets.
Sources (1)
- blog.google(opens in a new tab)Originalblog.google