Gemini 3.1 Pro First Look: Google Doubles the Reasoning Score That Matters

ℹ Editor's note, 23 September 2026: Google renamed NotebookLM to Gemini Notebook in July 2026. This is a dated record and keeps the name that was correct when it was written; the review carries the current name.

Launched: 19 February 2026 · Reviewed: 19 February 2026 · Score: 20/25 ⭐⭐⭐⭐ Very Good

Verdict in one line: A serious jump in reasoning that most Australians will get without paying a cent more — the quiet advantage of Google shipping models into products you already use.

📌 Correction — 20 September 2026. An earlier version of this note said Google's current flagship was Gemini 3.5. That was wrong. Google's model page still lists Gemini 3.1 Pro as its Pro-tier model, with 3.5 Pro "coming soon". The models shipped since are Gemini 3.8 Flash, Gemini 3.5 Flash-Lite and Gemini 3.1 Deep Think. This first look is therefore still current on the Pro tier. For our fully-scored assessment of Gemini, read the full Google Gemini review.

⚠️ STALENESS FLAG — 5 September 2026. This first look was published 19 February 2026 and is 198 days old. It has not been re-verified against the vendor since. AI products move fast: pricing, model versions, plan names and Australian availability may all have changed. Read the verdict below as a snapshot of February 2026, not as a current recommendation. (House rule: any review over six months old carries this flag until it is updated.)

What launched

On 19 February 2026 Google released Gemini 3.1 Pro, the new top model in the Gemini line. The headline claim: double the ARC-AGI-2 score of Gemini 3 Pro — the benchmark built specifically to resist memorisation and test genuine reasoning. It's rolling out to the Gemini API, Vertex AI, the Gemini app, and NotebookLM, with higher usage limits for Google AI Pro and Ultra subscribers.

What it does

ARC-AGI-2 is the number to watch because it can't be crammed for. Doubling it means the model is materially better at the kind of problem it hasn't seen before — which, in practice, is most of your problems. The other story is distribution. OpenAI and Anthropic launch models into chat windows; Google launches them into Gmail, Docs, NotebookLM, and search. A model upgrade here lands inside tools Australians already open every morning, with nothing to install and nothing new to learn.

Pricing for Australians

The model itself costs nothing extra — it slots into the existing Gemini app and Google's AU plans, which bill natively in AUD: AI Plus at A$7.99 a month, AI Pro at A$32.99, and AI Ultra from A$149.99. That is still the cleanest billing story of any major AI vendor here. The higher limits sit on the paid Pro and Ultra tiers; casual users will meet 3.1 Pro without touching their wallet.

Australian Fit

Available in Australia at launch through the app, AUD-billed plans, no waitlist games. Data processing remains offshore, as with every frontier model — keep that in mind for client-confidential work. 4/5.

Early read

Good: the reasoning jump is the real thing benchmarks exist to measure; day-one availability in Australia; lands inside products people already use; NotebookLM inclusion is quietly significant for research-heavy work. Not so good: "rolling out" means some surfaces get it before others and Google doesn't say which first; the best limits need a paid tier; and Google's model-naming churn makes it genuinely hard for a normal person to know which Gemini they're talking to.

Early verdict

If you're already in the Google ecosystem, you just got a smarter brain in everything for free — no action required. If you're choosing between assistants on reasoning-heavy work, 3.1 Pro puts Gemini firmly back in the conversation. Watch for our full review once the rollout settles.

First-look score — provisional until our full hands-on review.

Frequently asked questions

What is Gemini 3.1 Pro and what's new in it?

Gemini 3.1 Pro is Google's new top model in the Gemini line, released 19 February 2026. The headline claim: double the ARC-AGI-2 score of Gemini 3 Pro — the benchmark built specifically to resist memorisation and test genuine reasoning. It's rolling out to the Gemini API, Vertex AI, the Gemini app, and NotebookLM.

Is Gemini 3.1 Pro available in Australia?

Yes. It's available in Australia at launch through the Gemini app, on AUD-billed plans, with no waitlist games. One caveat: the rollout is staggered — some surfaces get it before others, and Google doesn't say which come first. Data processing remains offshore, so keep that in mind for client-confidential work.

Does Gemini 3.1 Pro cost extra for Australians?

No. The model slots into the existing Gemini app and Google's AU plans, which bill natively in AUD — AI Plus at A$7.99 a month, AI Pro at A$32.99, AI Ultra from A$149.99. That's the cleanest billing story of any major AI vendor here. Higher usage limits sit on the paid Google AI Pro and Ultra tiers; casual users will meet 3.1 Pro without touching their wallet.

How does Gemini 3.1 Pro compare with OpenAI and Anthropic's models?

Distribution is the difference. OpenAI and Anthropic launch models into chat windows; Google launches them into Gmail, Docs, NotebookLM, and search — tools Australians already open every morning, with nothing to install and nothing new to learn. On reasoning-heavy work, the ARC-AGI-2 jump puts Gemini firmly back in the conversation.

When is the full review of Gemini 3.1 Pro coming?

Once the rollout settles. Our first-look score is 20/25 — Very Good — and it stays provisional until the full hands-on review. Early knocks to watch: the staggered rollout, the best limits sitting on a paid tier, and Google's model-naming churn making it hard to know which Gemini you're talking to.

Sources

• Google — Gemini 3.1 Pro announcement: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/ (primary) [Wayback snapshot TBC]
• Google DeepMind — Gemini 3.1 Pro model card: https://deepmind.google/models/model-cards/gemini-3-1-pro/ (primary) [Wayback snapshot TBC]