Google DeepMind Ships Gemini 4 Neon, Faster and Cheaper Than Carbon
The Development
Google DeepMind released Gemini 4 Neon on November 18, its fourth model in seven weeks, positioned between Carbon and Hydrogen with a one-million-token multimodal context window. At its xhigh reasoning level, Neon scores 61 on the Artificial Analysis Intelligence Index against Carbon's 60, placing third behind OpenAI's GPT-6.2 Astra and Anthropic's Claude Fable 5.2. Output runs near 110 tokens per second versus Carbon's 57. List pricing is $3 per million input tokens and $15 per million output tokens, 25% below Carbon's regular rate but 50% above its promotional price, which runs through November. Per-task cost at xhigh is $6.44 versus $7.76 for Carbon. Neon also posted the top presentation-quality score on AA-Briefcase.
Our Take
The headline is speed and price, but the real signal is cadence. Four models in seven weeks means model advantage is a rental, not an asset. Note also that Neon's savings come entirely from list price, since it burns about 15% more output tokens per task than Carbon. That makes the economics a pricing decision Google can reverse, not an efficiency gain you can bank. Google is buying share with price while omitting the two models that outrank Neon from its own comparison table. Treat it as a strong, cheap workhorse for volume, not as the frontier. The strategic risk is lock-in to any single vendor's current price.
What Changed
Near-frontier multimodal reasoning now runs at roughly double the speed and at lower unit cost than Google's own flagship from weeks earlier. Strong document and presentation analysis arrives at that price point, making high-volume agentic content production and review economically viable at scale for the first time.
Marketing Impact
Creative operations and marketing operations gain the most: document-heavy workflows such as deck production, brief analysis, asset QA and report generation get roughly half the latency and lower per-task cost. Teams running agentic content pipelines can rerun cost models now, since Neon's high setting matches Carbon's quality at half the per-task price.
Competitive Implication
In-house teams with model-agnostic pipelines gain structural advantage, because they can swap in Neon within days and capture the price drop. Agencies and martech vendors that resell AI output at fixed per-seat or retainer pricing become vulnerable as unit costs fall and clients demand the savings. Anthropic and OpenAI face renewed price pressure at the mid-tier.
Strategic Outlook
Expect Anthropic and OpenAI to answer with mid-tier price cuts or faster variants within weeks. Carbon's promotional pricing ends this month, so Google's effective price floor rises in December. Prices will keep falling in 2027, but the release cadence will make quarterly model re-evaluation a standing operational requirement, not an occasional project.
The Exploit
Action Item
Marketing operations leads running generative or agentic pipelines should benchmark Neon at the high reasoning setting against their current model on their own briefs and documents before Carbon's promotional pricing expires on November 30, then lock a model-swap review into the Q1 2027 plan.