Google AI Studio all models available: Gemini 3 general availability, Gemini 2.5 production tiers, and legacy options for early 2026

Graziano Stefanelli
2 hours ago
3 min read

Google AI Studio has evolved into a central access point for multiple generations of Gemini models, reflecting Google’s shift toward rolling availability rather than isolated model launches.

What users see inside AI Studio today is no longer limited to experimental previews or transitional builds, but a layered mix of generally available models, stable production tiers, and legacy options kept for compatibility and benchmarking.

Here we share how the current model lineup in Google AI Studio is structured, which Gemini families are actually available, and how this setup reveals Google’s broader strategy as usage consolidates into early 2026.

··········

Gemini 3 models are generally available inside Google AI Studio.

Gemini 3 models are no longer restricted to preview-only exposure within AI Studio.

Both Gemini 3 Flash and Gemini 3 Pro appear as selectable options with direct access for experimentation.

Gemini 3 Flash has become the default model across several Google surfaces, reinforcing its status as a generally available, production-ready option rather than an experimental build.

Gemini 3 Pro targets higher-reasoning workloads and agent-oriented tasks while sharing the same multimodal foundation.

··········

·····

Gemini 3 models available in Google AI Studio

Model	Primary role	Positioning
Gemini 3 Flash	Fast general-purpose reasoning	Default, GA
Gemini 3 Pro	Advanced reasoning and agents	GA

··········

Gemini 2.5 remains the stable production backbone for many workflows.

Despite the availability of Gemini 3, the Gemini 2.5 family continues to play a central role in AI Studio.

These models are widely used for prompt engineering, evaluation, and workloads that require predictable latency and mature behavior.

Gemini 2.5 models also retain very large context windows and well-documented token limits, making them reliable for long-form and multimodal tasks.

For teams prioritizing stability over frontier behavior, Gemini 2.5 remains a preferred choice.

··········

·····

Gemini 2.5 models commonly visible in AI Studio

Variant	Focus
Gemini 2.5 Pro	Deep reasoning and long-context tasks
Gemini 2.5 Flash	Speed-balanced general use
Gemini 2.5 Flash-Lite	Low-latency, cost-efficient tasks
Gemini 2.5 Flash Image	Image-first multimodal prompts

··········

Legacy Gemini 2.0 models remain accessible for compatibility and cost control.

Google AI Studio still references Gemini 2.0 Flash, Flash-Lite, and Flash Image models.

These options exist primarily to support older prompt designs, regression testing, and budget-sensitive experimentation.

They are not the focus of active feature development, but their presence allows developers to compare behavior across generations.

This continuity reduces friction during migration between model families.

··········

Modality-specific Gemini endpoints extend AI Studio beyond text chat.

AI Studio surfaces specialized Gemini endpoints when prompts involve audio or image-heavy tasks.

Audio-native models and text-to-speech variants are available as selectable targets for speech-focused workflows.

These models are not always visible in the default model list but appear contextually based on project configuration.

This reflects Google’s move toward modality-driven model selection rather than a single chat-centric interface.

··········

·····

Examples of modality-focused models in AI Studio

Model type	Primary modality
Flash Image variants	Vision-heavy prompts
Native Audio variants	Speech input and output
TTS preview endpoints	Text-to-speech generation

··········

Model names represent capability classes rather than static versions.

In Google AI Studio, a model label such as “Gemini 3 Pro” does not map to a fixed internal build.

Google updates underlying weights and behavior continuously without changing the visible name.

As a result, model behavior can improve or shift over time even when the selection remains unchanged.

This rolling-update approach favors iteration speed over rigid versioning.

··········

AI Studio intentionally excludes enterprise-only and private Gemini variants.

Not all Gemini models are exposed through AI Studio.

Enterprise-tuned, compliance-focused, or privately fine-tuned models remain exclusive to Vertex AI and internal deployments.

AI Studio focuses on accessibility and experimentation rather than production governance.

This separation keeps the platform lightweight while guiding large-scale workloads toward Google Cloud.

··········

Model availability in AI Studio evolves quietly and continuously.

New models may appear without formal announcements.

Older models may be deprecated with minimal notice.

Availability can differ by region, account type, and rollout stage.

Users should treat AI Studio as a dynamic environment rather than a static catalog.

··········

Choosing a Gemini model in AI Studio depends on workflow intent.

Gemini 3 Flash suits fast, interactive tasks and default experimentation.

Gemini 3 Pro is appropriate for complex reasoning and agentic workflows.

Gemini 2.5 Pro remains a strong choice for long-context stability and repeatable evaluations.

Legacy Gemini 2.0 models serve as baselines and cost-saving options.

Selecting models based on capability needs rather than version numbers leads to more reliable outcomes.

··········

DATA STUDIOS

··········

[datastudios.org]