Frontier token prices: Opus drops to $5, Gemini 3 Flash lands at $0.50

Two price points in the past month change the arithmetic for anyone running models at volume. Anthropic priced Claude Opus 4.5, released November 24, at $5 per million input tokens and $25 per million output tokens ^[2]. Opus 4 launched in May alongside Sonnet 4, at $15 and $75 ^[1]. On December 17 Google released Gemini 3 Flash in preview at $0.50 per million input tokens and $3 per million output tokens ^[4].

Prices side by side

List prices only; vendors run different tiers, batch discounts and caching rules.

Why list price understates the change

Both vendors pair price with efficiency claims. Anthropic says Opus 4.5 at medium effort matches Sonnet 4.5's best SWE-bench Verified score while using 76% fewer output tokens ^[2]. Google says Gemini 3 Flash scores 78% on SWE-bench Verified, ahead of Gemini 3 Pro, and uses about 30% fewer tokens on average than Gemini 2.5 Pro on typical traffic ^[4]. Both are vendor claims on vendor-chosen tests. If they hold on your workload, cost per completed task falls further than the price list alone suggests. If they do not, the discount is illusory.

A drop on one flagship does not mean every model is cheaper. Premium tiers and long-context pricing still apply, and preview pricing can change at general availability.

The practical implication: rerun your cost-per-task estimates on current prices and test a cheaper tier on your own evaluation set before assuming a flagship is required.