# Gemini 3.1 Pro

Gemini 3.1 Pro is an AI model released by Google on Feb 19 2026. Tracked results include 37% on BullshitBench v2, 94.3% on GPQA Diamond and 80.6% on SWE-Bench Verified.

## Facts

| Field | Value |
| --- | --- |
| Model | Gemini 3.1 Pro |
| Developer | Google |
| Release date | Thursday, Feb 19 2026 |
| Licensing | Proprietary |

## API pricing

All rates in USD per 1,000,000 tokens, pay-as-you-go.

| Tier | Input | Cached input | Output |
| --- | --- | --- | --- |
| Prompts up to 200K tokens | $2.00 | $0.20 | $12.00 |
| Prompts over 200K tokens | $4.00 | $0.40 | $18.00 |

Output prices include reasoning tokens.
Verified August 18, 2026 against the first-party source: https://ai.google.dev/gemini-api/docs/pricing

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| BullshitBench v2 | 37% | [BullshitBench](https://github.com/petergpt/bullshit-benchmark) | Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. |
| Gray Swan IPI (k = 1) | 14.2% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. Gray Swan's indirect prompt injection benchmark measures how often such an attack succeeds when the attacker gets a single try. Lower is better. |
| Gray Swan IPI (k = 10) | 45.7% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 10 tries and counts an attack as successful if any of them works. Lower is better. |
| Gray Swan IPI (k = 15) | 49.2% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 15 tries and counts an attack as successful if any of them works. Lower is better. |
| ProgramBench | 0% | [ProgramBench](https://programbench.com/), retrieved 2026-09-11 | The AI receives a working program and its documentation, then builds a replacement from scratch without the original source code, internet access or decompilation. The score is the percentage of 200 programs that pass every behavioral test. We record each model's best published mini-SWE-agent result, including higher reasoning efforts where available. Partial test-pass rates and almost-solved programs do not count toward this score. Equal scores share a rank here; the official board also uses partial progress to break ties. Higher is better. |
| SWE-Bench Pro | 54.2% | Lab | Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. |
| SWE-Bench Verified | 80.6% | Lab | Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. |
| DeepSWE 1.1 | 12% | Lab | Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better. |
| MLE-Bench | 42.6% | Lab | Can the AI do the work of a machine-learning engineer? It competes in real Kaggle competitions — building, training, and tuning models end to end — and the score reflects how well it places. Higher is better. |
| Next.js Evals | 58% | [Next.js Evals](https://nextjs.org/evals) | Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better. |
| Terminal-Bench 2.1 | 70.3% | Lab | Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better. |
| Terminal-Bench 2.0 | 68.5% | Lab | Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? (Version 2.0 of the test.) Higher is better. |
| MCP Atlas | 78.2% | Lab | Can the AI chain together many tools and steps to complete one bigger task, rather than doing just a single thing? Higher is better. |
| Toolathlon | 48.8% | Lab | Tests how well the AI uses everyday real-world tools and apps to get things done. Higher is better. |
| BrowseComp | 85.9% | Lab | Can the AI browse the web and track down hard-to-find answers? Higher is better. |
| Humanity's Last Exam (no tools) | 44.4% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. |
| Humanity's Last Exam (with tools) | 51.4% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. |
| ARC-AGI-2 | 77.1% | Lab | Puzzle-style tests of abstract reasoning and pattern-finding — the kind of thing people find easy but AIs often struggle with. Higher is better. |
| FrontierMath (Tier 1–3) | 36.9% | Lab | Very hard, research-level math problems. Tiers 1–3 are the (still extremely difficult) lower tiers. Higher is better. |
| FrontierMath (Tier 4) | 16.7% | Lab | Very hard, research-level math problems. Tier 4 is the hardest — close to what professional research mathematicians tackle. Higher is better. |
| GPQA Diamond | 94.3% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. |
| OSWorld-Verified | 76.2% | Lab | Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Higher is better. |
| Finance Agent v2 | 43% | Lab | Tests the AI on real financial-analysis work, like digging through reports and making sound decisions. Higher is better. |
| GDPval-AA | 1314 | Lab | Measures how well the AI does economically valuable knowledge work, judged against human experts. Shown as a rating (like a chess Elo) — higher is better. |
| GDPval-AA v2 | 965 | Lab | economically valuable knowledge work (v2, re-based Elo) |
| GDPval (win/tie rate) | 67.3% | Lab | How often the AI's work matches or beats a human expert's on real knowledge-work tasks. Higher is better. |
| CharXiv Reasoning | 83.3% | Lab | Can the AI read and reason about complex charts and figures, not just text? Higher is better. |
| MMMU-Pro | 80.5% | Lab | A tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better. |
| Blueprint-Bench 2 | 26.5% | Lab | Can the AI reason about space and layout — for example, understanding a floor plan or blueprint? Higher is better. |
| MRCR v2 (8-needle) (128k average) | 84.9% | Lab | Tests whether the AI can find specific details buried inside a very long document (around 128k tokens — roughly a long book). Higher is better. |
| MRCR v2 (8-needle) (1M pointwise) | 26.3% | Lab | Tests whether the AI can find specific details buried inside an enormous document (around 1 million tokens — many books). Higher is better. |

## About Gemini 3.1 Pro

Gemini 3.1 Pro, released February 19, 2026, sharpened the 3-series flagship into a reasoning specialist: 94.3% on GPQA Diamond and 77.1% on ARC-AGI-2 were both the top published scores at release, alongside 85.9% on BrowseComp for agentic web research and 84.9% on MRCR v2 128K long-context recall.

It landed in the same fortnight as Anthropic's Claude Sonnet 4.6 and xAI's Grok 4.20 Beta, in one of 2026's densest release windows. The 3.1 generation filled out below it with Gemini 3.1 Flash-Lite in early March, before Google's attention shifted to the 3.5 Flash line in May 2026.

## Questions and answers

### When was Gemini 3.1 Pro released?

Gemini 3.1 Pro was released by Google on Thursday, Feb 19 2026.

### Who made Gemini 3.1 Pro?

Gemini 3.1 Pro was built by Google. Builds the Gemini family of models through Google DeepMind. Integrates AI across Google products.

### How much does Gemini 3.1 Pro cost?

Gemini 3.1 Pro costs $2.00 per million input tokens and $12.00 per million output tokens through the Google API. Cached input is $0.20 per million tokens. Those are the rates for the “Prompts up to 200K tokens” tier; 1 other pricing tier is published for this model. Output prices include reasoning tokens. Rates are pay-as-you-go API prices verified against Google's published pricing on August 18, 2026.

### What benchmark scores did Gemini 3.1 Pro get?

Gemini 3.1 Pro reports 31 tracked benchmark scores — BullshitBench v2: 37%; Gray Swan IPI (k = 1): 14.2%; Gray Swan IPI (k = 10): 45.7%; Gray Swan IPI (k = 15): 49.2%; ProgramBench: 0%; SWE-Bench Pro: 54.2%; SWE-Bench Verified: 80.6%; DeepSWE 1.1: 12%; MLE-Bench: 42.6%; Next.js Evals: 58%; Terminal-Bench 2.1: 70.3%; Terminal-Bench 2.0: 68.5%; MCP Atlas: 78.2%; Toolathlon: 48.8%; BrowseComp: 85.9%; Humanity's Last Exam (no tools): 44.4%; Humanity's Last Exam (with tools): 51.4%; ARC-AGI-2: 77.1%; FrontierMath (Tier 1–3): 36.9%; FrontierMath (Tier 4): 16.7%; GPQA Diamond: 94.3%; OSWorld-Verified: 76.2%; Finance Agent v2: 43%; GDPval-AA: 1314; GDPval-AA v2: 965; GDPval (win/tie rate): 67.3%; CharXiv Reasoning: 83.3%; MMMU-Pro: 80.5%; Blueprint-Bench 2: 26.5%; MRCR v2 (8-needle) (128k average): 84.9%; MRCR v2 (8-needle) (1M pointwise): 26.3%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### Is Gemini 3.1 Pro open source?

No. Gemini 3.1 Pro is a proprietary model. The weights are not published — it is available only through the provider's own API, apps, or partner platforms.

### What came before and after Gemini 3.1 Pro?

Google's previous tracked release was Gemini 3.0 Flash on Dec 17 2025, 64 days earlier. It was followed by Gemini 3.1 Flash-Lite on Mar 3 2026.


---

Canonical page: https://aireleasetracker.com/model/google/gemini-3.1-pro
Full dataset: https://aireleasetracker.com/llms-full.txt · JSON: https://aireleasetracker.com/models.json
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.