Models & Releases

Google's New AI Codes for a Fraction of Claude's Price

Cheaper AI that writes software could mean lower prices on apps you use.

Deep Dive

Google quietly released a new AI model called Gemini 3.8 Flash, built for what the industry calls "agentic coding" — AI that doesn't just suggest code, but writes it, tests it, and fixes its own mistakes. The headline is the price. Google charges $0.75 for every million words the AI reads and $3.75 for every million it writes. Claude Opus 5, made by rival Anthropic, charges $5 and $25 for the same work. That's roughly six times more expensive per word. Gemini is also much quicker: about 305 words per second, compared with Claude's 56.

Why should you care if you don't write code? Because software companies pay these bills, and those bills shape what they can afford to build and how fast. A tool that's six times cheaper can mean smaller startups ship products sooner, bigger companies automate more of the boring work, and everyday apps get updated faster. Some routine programming jobs may shift from "write this" to "check what the AI wrote." Historically, cheaper software production has meant more software, not less.

The catch is that cheaper per word doesn't always mean cheaper per task. Gemini tends to work in small steps and double-check itself, which burns through more words. Independent testers at Artificial Analysis still give Claude a higher overall intelligence score: 63 versus Gemini's 59. Claude also does better on some tricky problem-solving tests. So Gemini wins on value, while Claude wins on raw quality for the hardest jobs.

Most professional teams will probably use both — the cheap, fast one for routine work and the premium one for the hardest problems. Think of it like choosing between a budget courier and a specialist. The real story isn't one model beating another. It's that the cost of getting AI to do real work keeps falling, and that drop eventually reaches you in the form of cheaper, faster, more capable software.

Key Points
  • Gemini 3.8 Flash costs about $0.75 per million words read and $3.75 per million written — roughly six times cheaper than Claude Opus 5's $5 and $25.
  • It's dramatically faster too, replying at about 305 words per second versus Claude's 56, and it can handle text, images, audio and video.
  • Independent testers still score Claude slightly higher on overall smartness (63 vs 59), so the cheap option isn't automatically the best one.

Why It Matters

Companies building apps with AI could spend far less — savings that may reach you as cheaper, faster software.

📬 Get the top 10 AI stories daily