Gemini 3.1 Pro: One Million Tokens of Context and 77% on ARC-AGI-2
One Million Tokens: Why It Matters
Google continues to bet on long context. Gemini 3.1 Pro retains its record-setting window of 1 million tokens — roughly 700,000 words, or several full-length books.
For practical purposes, this means:
- Loading an entire codebase of a medium-sized project in a single request
- Analyzing annual financial reports without losing context
- Working with long conversation histories and documentation
- Processing multi-hour transcriptions of meetings and calls
ARC-AGI-2: Testing Abstract Reasoning
The ARC-AGI-2 benchmark tests a model’s ability to perform abstract reasoning — tasks that a child can easily solve but that stump most AI systems.
Gemini 3.1 Pro scored 77% on ARC-AGI-2, one of the best results among commercial models:
| Model | ARC-AGI-2 |
|---|---|
| Claude Sonnet 4.6 | 79.2% |
| Gemini 3.1 Pro | 77.0% |
| GPT-5.3 | 74.5% |
| DeepSeek V3 | 71.3% |
Market Share Is Growing
According to analysts, Google’s share of the LLM API market has grown from 12% to 18% over the last six months. Key reasons:
Pricing
Google offers some of the most competitive prices:
- Input: $3.50 / 1M tokens
- Output: $10.50 / 1M tokens
- For context >128K: $7 / $21 per million
Google Cloud Ecosystem
Integration with Vertex AI, BigQuery, and other Google Cloud services makes Gemini attractive to enterprise clients already using Google’s cloud infrastructure.
Multimodality
Gemini 3.1 Pro natively supports:
- Text, images, audio, and video
- Code generation and analysis
- Working with tables and structured data
What This Means for Traders
The 1M token long context opens up interesting possibilities:
- Loading entire trade histories over extended periods for pattern analysis
- Simultaneously analyzing multiple company financial reports
- Processing a full day’s news feed without missing important details
That said, the quality of working with information at the beginning and end of a long context may differ — the so-called “lost in the middle” problem.
Google is firmly securing its place among the top three AI market leaders, and Gemini 3.1 Pro is a strong argument in this race.
Discussion
Join the discussion in our Telegram chat!