Gemini 3.1 Pro Available in Gemini CLI

Gemini 3.1 Pro Arrives in Gemini CLI

Google released Gemini 3.1 Pro on February 19, 2026, and Gemini CLI users gained immediate access to the model via the gemini-3.1-pro-preview API identifier. For developers who rely on Gemini CLI as their daily coding assistant, this represents one of the most significant capability upgrades since the tool launched.

Breakthrough Reasoning Performance

The headline improvement in Gemini 3.1 Pro is reasoning. On the ARC-AGI-2 benchmark — widely considered the most demanding test of generalizable intelligence — Gemini 3.1 Pro scored 77.1%, more than doubling the 31.1% achieved by Gemini 3 Pro. The model also reached 80.6% on SWE-Bench Verified, the standard benchmark for real-world software engineering tasks, and 94.3% on GPQA Diamond, which evaluates graduate-level scientific reasoning.

These numbers matter to Gemini CLI users specifically because the tool's value is directly tied to how well the underlying model can reason through complex code, understand large repositories, and produce correct multi-step solutions.

Adjustable Thinking Levels

One of Gemini 3.1 Pro's most practical features for CLI use is its three-tier thinking system. Where previous Gemini 3 Pro offered a single reasoning mode, Gemini 3.1 Pro introduces low, medium, and high settings. At the high setting, the model behaves as a condensed version of Gemini Deep Think — Google's specialized reasoning system previously available only to Ultra subscribers. This means Gemini CLI users can dial in the appropriate level of reasoning depth for a given task, trading latency for accuracy when working through particularly difficult engineering problems.

Context Window and Developer Capabilities

Gemini 3.1 Pro supports a 1 million token context window with up to 64,000 tokens of output — sufficient to load an entire mid-sized codebase in a single session. Multimodal reasoning spans text, images, audio, video, and code. Agentic workflows, algorithm design, and real-world ML engineering tasks are explicitly highlighted in the model card as target use cases, all of which map directly to how Gemini CLI is used in practice.

Same Price, Major Upgrade

For developers already using Gemini 3 Pro via the Gemini API, Gemini 3.1 Pro carries identical pricing — making the performance jump a no-cost upgrade. The model is accessible through Google AI Studio, the Gemini API, Gemini CLI, Vertex AI, Gemini Enterprise, and Android Studio.

Bug Fixes and API Updates

Alongside the Gemini 3.1 Pro launch, Google's February 18 API changelog noted that several Gemini 2.0 models will be retired on June 1, 2026, including gemini-2.0-flash, gemini-2.0-flash-001, gemini-2.0-flash-lite, and gemini-2.0-flash-lite-001. Developers using these model identifiers directly should plan migrations to Gemini 3-series models before that date.