Google Unleashes Gemini 4 as Four Models With Four Price Tags
Google's Gemini 4 isn't one model — it's four. Argon, Carbon, Hydrogen and Neon all share a 1M-token context window but split on reasoning depth, speed and price, and Carbon's rates double after November 30. Developers now pick both a model and how hard it thinks.
- Original source: cryptobriefing.com.
- The source published it on Nov 19, 2026 at 8:41 PM.
- No second outlet has reported it yet. Treat it as a single-source claim.
Google's Gemini 4 family is out, and it arrives as four models rather than one: Argon, Carbon, Hydrogen and Neon. All four share a unified 1M-token context window, enough to hold a large pile of text, code or documents in working memory at once. The differences show up in reasoning depth, execution speed and price, with Google slotting each variant into a different slice of work running from casual tinkering to enterprise jobs.
Pricing is quoted per million tokens and split between input and output, with output costing five times input across the whole lineup. That creates a short-lived quirk: through November 30, Carbon is cheaper per token than Neon. After the cutoff, Carbon's rates double and it becomes the priciest model of the group, with Carbon and Argon moving to $4 input and $20 output. Hydrogen sits at $1.50 input and $7.50 output, and its output rate is exactly half of Neon's — a gap that adds up fast for text-heavy workloads.
Each model takes an adjustable reasoning effort dial: low, medium, high and, in some cases, xhigh. More thinking generally means more computation per request, which usually translates into more output tokens and a longer wait for an answer. Carbon is tuned for medium effort and performs best of the group at low effort, while Neon and Hydrogen are optimized for high effort — the setting where they start closing the gap and scale better.
On hard terminal-style tasks, Carbon reportedly hits approximately 65% on Terminal-Bench 4.0. In coding evaluations, Neon matches Carbon's results while burning less time and money, according to Crypto Briefing. Carbon is pitched as the pick for complex problems where a wrong answer costs more than a slow one; Neon chases intricate work with an eye on speed and cost; Hydrogen handles everyday jobs fast at a lower price. On consumer surfaces, Carbon is restricted to the Ultra tier.
For developers, the choice now runs on two axes: which model, and how hard it should think. A team that can tolerate high reasoning effort may find Neon delivering Carbon-level coding results for less time and money, per the source. A team that needs strong answers fast with minimal thinking overhead has a clearer case for Carbon — especially before November 30, while its rates sit below Neon's. Workloads that defaulted to Carbon may get rerouted to Neon or Hydrogen once the higher rates kick in.
TopicsGoogle · Gemini 4 · Carbon · Argon · Hydrogen · Neon · Terminal-Bench 4.0
Source: cryptobriefing.com
Written by AI from verified reporting.
More stories
- Tech & AI cryptobriefing.com 0 viewsAMD takes record 35.5% of server CPUs, Intel down to 64.5%
- Tech & AI cryptobriefing.com 1 viewGoogle slides an AI chat sidebar onto its famously blank homepage
- Tech & AI cryptobriefing.com 3 viewsAlibaba, Tencent, Baidu and BYD seek US licenses for advanced chips
- Tech & AI x.com 3 viewsMeta's Muse AI agent reportedly read private messages, per Reuters

