Logan Kilpatrick
@OfficialLoganK
Gemini 3.5 Flash has made huge progress from 3.1 Pro on GDPval, Flash is competing at the frontier, post training going strong :) https://t.co/XW1cyiv1Ru
33retweets836likes
View on XGoogle Gemini 3.5 Flash Surpasses Previous Pro Model on Key Benchmark
Google confirmed that Gemini 3.5 Flash has surpassed the performance of the previous Gemini 3.1 Pro on the GDPval benchmark. Logan Kilpatrick, who leads Google AI Studio, noted that the model's post-training (refining a model for specific behaviors after initial training) is yielding frontier-level results.
- Benchmark
- GDPval
- Comparison
- Surpassed Gemini 3.1 Pro
- Availability
- Gemini API and Google AI Studio
- Future update
- Efficiency gains in next revision
This milestone marks a significant shift in the Gemini hierarchy, where the high-speed tier now matches the reasoning depth of previous mid-tier models. It validates the Gemini 3.5 Flash launch strategy of delivering flagship intelligence within a high-efficiency architecture, mirroring its top performance on the Zapier Automation Benchmark.
You can now migrate legacy workflows built for Gemini 3.1 Pro to the faster Gemini 3.5 Flash without sacrificing quality, as seen on the Arena.ai coding leaderboards. The model is available via the Gemini API and Google AI Studio. This shift allows you to maintain frontier-level reasoning while benefiting from the lower latency and higher throughput of the Flash architecture.
Still wondering? A few quick answers below.
Gemini 3.5 Flash has officially surpassed the performance of the previous Gemini 3.1 Pro model on the GDPval benchmark. This shift means that Google's high-speed, efficient model tier is now capable of matching or exceeding the reasoning depth and quality of its previous-generation mid-tier professional model.
GDPval is a performance evaluation metric used by Google to measure the progress and capability of its AI models. Recent results on this benchmark confirm that Gemini 3.5 Flash is now competing at the frontier level, demonstrating that the model's post-training phase has significantly improved its intelligence and accuracy.
Gemini 3.5 Flash is available to developers and enterprise users through the Gemini API and Google AI Studio. Because it has overtaken the capabilities of the older Gemini 3.1 Pro, users can migrate their existing workflows to this newer model to gain higher speeds without losing performance quality.
Google is currently working on the next revision of the Gemini 3.5 Flash model, which is expected to introduce further efficiency gains. These updates aim to improve the model's performance and speed even further, maintaining its position as a high-speed workhorse for complex autonomous and agentic tasks.
The Flash label represents Google's workhorse model tier, designed for high-speed and cost-effective performance. According to Google DeepMind leadership, the definition of Flash is dynamic and changes over time as user use cases evolve, ensuring that the most efficient tier always provides competitive frontier-level intelligence.
