Gemini 3.1

Overview

Gemini 3.1 is a Google Gemini model variant referenced across prototyping, benchmark evaluation, and AI-assisted design workflows. In the newsletter mentions, it appears in two especially notable contexts: as the top AI model measured on the ARC AGI 3 benchmark at a still very low 0.37% score versus humans, and as a model capable of generating high-fidelity visual design outputs inside Figma when given complex instructions and reference images. It is also tied to Google AI Studio’s updated prototype-first workflow, where it was used to help recreate and simplify an existing product interface in a matter of minutes.

For AI Product Managers, Gemini 3.1 matters because it sits at the intersection of product prototyping, multimodal UI generation, and model capability assessment. The mentions suggest it is useful not just as a chat model, but as a practical building block for turning prompts and visual references into product concepts, while also illustrating the gap between polished output generation and deeper general reasoning performance. That combination makes it relevant for PMs evaluating where frontier models are strong today and where expectations should remain grounded.

Key Developments

• 2026-02-21 — Gemini 3.1 was highlighted alongside a major Google AI Studio full-stack update. In a tutorial context, it was used in a five-step prototype-first workflow to replicate and simplify the AI Studio UI, reportedly producing a redesigned interface in roughly 141 seconds. The workflow also emphasized support for in-tool servers, databases, and multiplayer features.
• 2026-03-27 — Gemini 3.1 was cited as the top AI model on the ARC AGI 3 benchmark, but with only 0.37% performance against a 100% human baseline. This mention framed the model as state-of-the-art within the benchmark while underscoring how far current AI still is from human-level general problem solving.
• 2026-04-13 — Gemini 3.1, together with Gemini 3.0, was described in Figma’s AI-driven design workflow as capable of producing high-fidelity visual design outputs on the Figma canvas when prompted with complex instructions and reference images. This positioned the model as part of a modern design generation loop that combines AI exploration with direct manipulation in Figma.

Relevance to AI PMs

1. Accelerates prototype-first product discovery AI PMs can use Gemini 3.1-style workflows to move from requirements or interface references to functional UI concepts quickly. That is useful for validating flows, testing visual directions, and reducing time between idea and stakeholder review.

2. Supports multimodal design collaboration
The Figma mention suggests Gemini 3.1 can work from both text instructions and reference imagery. For PMs, that means clearer collaboration across product, design, and engineering teams: product intent can be expressed as prompts, while visual examples anchor the output.

3. Helps calibrate model selection and expectations
The contrast between strong design-generation output and weak ARC AGI 3 performance is tactically important. PMs should evaluate Gemini 3.1 based on the job to be done: it may be effective for high-fidelity interface generation and prototyping, but not evidence of robust general reasoning across unfamiliar tasks.

Related

• Google DeepMind — Closely associated with the Gemini family and the broader advancement of Gemini model capabilities.
• Google AI Studio — A key product context for Gemini 3.1, especially in prototype-first and full-stack app-building workflows.
• Figma — Gemini 3.1 was referenced as part of an AI-assisted design generation workflow inside Figma’s canvas environment.
• Gemini 3.0 / gemini-30 — Mentioned alongside Gemini 3.1 in design-generation use cases, suggesting comparable or adjacent capability in visual output tasks.
• ARC AGI 3 / arc-agi-3 — Benchmark where Gemini 3.1 was cited as the top AI model despite very low absolute performance relative to humans.
• Peter Yang — Featured in coverage of the Google AI Studio and Gemini 3.1 prototyping workflow tutorial.
• Dylan Field — Discussed Figma’s AI-driven design workflow in which Gemini 3.1 appeared as a model for high-fidelity visual output.
• Gemini app — Related as part of the broader Gemini product ecosystem, though the newsletter mentions here focus more on model and tooling workflows than the consumer app.