Google has released Gemini 3.1 Pro at the same $2 input / $12 output pricing as its predecessor Gemini 3 Pro — no price increase despite a major performance leap. The model scores 77.1% on ARC-AGI-2, more than doubling Gemini 3 Pro's reasoning performance and making it the best price-to-performance ratio among any closed frontier model this month.
For coding tasks specifically, Gemini 3.1 Pro shows marked improvements in multi-file reasoning, instruction following, and the ability to work within existing codebases without hallucinating APIs or imports. It is now the default model in Gemini Code Assist and available through Gemini CLI, which received its own update to v0.33 with HTTP authentication for remote agents and Plan Mode research subagents.
Gemini CLI itself remains one of the most accessible entry points for AI coding. It is completely free for individual developers using a personal Google account, with access to Gemini 2.5 Pro, a 1 million token context window, and allowances of 60 requests per minute and 1,000 per day. The combination of a free, powerful CLI tool and the new Gemini 3.1 Pro model makes Google's offering increasingly competitive with paid alternatives.
