What will Gemini 3.5 Pro's standard API OUTPUT price be, in USD per 1,000,000 tokens at the ≤200k-context tier, at public release?

As of mid-June 2026, Gemini 3.5 Pro is announced but not yet released, with general availability expected soon featuring a 2M context window and integrated Deep Think [aiweekly.co], techtimes.com, aiweekly.co. The central uncertainty for its standard output pricing revolves around how Google will balance its established 'price-performance leadership' strategy against evident generational price inflation and the compute costs of new reasoning capabilities.

The Floor and Generational Inflation ($12–$18) A firm floor exists at $12, the output price of the previous-generation Gemini 3.1 Pro for the ≤200k tier ai.google.dev, ai.google.dev. However, carrying this price forward is highly unlikely due to internal tier compression. Google priced Gemini 3.5 Flash at $9 per 1M output tokens ai.google.dev—a sharp ~3x increase over the prior Flash generation simonwillison.net. A flat $12 Pro model would sit at a commercially awkward 1.33x multiple of Flash. To restore a standard 2x–3x gap between the Flash and Pro tiers, the price must rise.

The Strategic Sweet Spot ($24–$39) Google explicitly markets Gemini's frontier-level capabilities as being available at 'less than half the cost' of competitors blog.google. With GPT-5.4 priced at $60 and Claude Opus at ~$75 for output, Google has significant runway to raise the price of 3.5 Pro while strictly preserving its brand positioning. A price in the mid-twenties to mid-thirties—reflected in the $24.33 median and $39.33 75th percentile—perfectly harmonizes these constraints. It establishes a healthy multiple over 3.5 Flash while aggressively undercutting OpenAI and Anthropic by 50% or more.

The Premium Repricing Tail ($58+) A secondary-source rumor suggests Gemini 3.5 Pro could launch at ~$60 per 1M output tokens, framed as roughly '10x Flash' aiweekly.co, techtimes.com, aiweekly.co. While this is a weakly sourced, speculative estimate rather than a verified leak techtimes.com, aiweekly.co, it presents a credible alternate scenario where Google abandons aggressive undercutting to ration and monetize scarce Deep Think reasoning capacity. The 90th percentile stretches to $58.33 to capture the risk of this full premium repricing.