Loading...

Google DeepMind has released Gemini 4 Neon, its fourth model in seven weeks, which is nearly twice as fast and significantly cheaper than its predecessor, Gemini 4 Carbon. This release marks a critical inflection point for small and mid-sized companies, as the cost-per-task for high-level AI reasoning has dropped below the $7 threshold, making autonomous agent workflows economically viable for teams with limited budgets.
According to The Decoder, Gemini 4 Neon is positioned between Carbon and Hydrogen in Google’s lineup and features a one-million-token context window for text, images, audio, and video. At its highest reasoning level, "xhigh," Neon scores 61 points on the Intelligence Index, edging past Carbon’s 60 points. While it trails OpenAI’s GPT-6.2 Astra and Anthropic’s Claude Fable 5.2, Google omits these competitors from its own comparison tables.
Pricing is the headline feature. Through Google AI Studio and Gemini for Business, one million input tokens costs $3 (or $0.15 when cached), and one million output tokens costs $15. This is 25% below Carbon’s regular price, though 50% above Carbon’s promotional rate, which remains active through the end of November. Artificial Analysis reports that per-task costs for Neon at xhigh are $6.44, roughly 17% lower than Carbon’s undiscounted $7.76. Notably, Neon uses about 15% more output tokens per task, meaning the savings stem entirely from the lower per-token list price. At the "high" reasoning level, Neon matches Carbon’s intelligence score while cutting per-task costs in half. Speed has also improved dramatically, with processing times dropping from roughly 36 minutes to 21 minutes as Neon generates 110 tokens per second compared to Carbon’s 57.
For developers, Neon is particularly competitive on the Coding Agent Index. Inside Google Antigravity, Neon at xhigh scores 59 points, beating Carbon (55) and Claude Opus 5.5 (58). An average coding task costs $6.86 with Neon, compared to $12.61 for Opus and $15.47 for Fable. In the SWE-sweep test, the Antigravity environment triples Neon’s hit rate to 39.0 percent. Additionally, Artificial Analysis measured large gains in visual processing, with Neon leading the AA-Briefcase test for presentation quality with 1771 Elo points.
The primary barrier to adopting AI agents in small businesses has never been capability; it has been the unpredictable and often prohibitive cost of inference. For a small team, the difference between a $15.47 and a $6.86 per-task cost is the difference between a pilot project and a scalable operational tool. Neon’s ability to match or exceed the coding performance of more expensive competitors like Claude Opus 5.5 at nearly half the price changes the unit economics of AI-driven development. Furthermore, the 40% reduction in latency (from 36 to 21 minutes) is not just a technicality; it transforms AI from a background batch process into an interactive tool that can be integrated into real-time workflows. For small companies, this speed allows for iterative testing and immediate feedback loops, which are essential for building reliable internal tools without a dedicated DevOps team.
Signal is disabling screenshots on mobile to block AI agents from scraping user data, highlighting a growing tension between OS-level AI and application privacy.

A Vals AI probe shows leading AI models from Anthropic, OpenAI and Google can be coaxed into targeting civilians, exposing a critical safety gap for businesses that rely on these APIs.

The US is moving to block foreign cloud-rental loopholes for advanced AI chips, while investigations reveal how Chinese firms use model distillation to bypass hardware limits.

Apple’s new iOS app invites up to 200,000 U.S. users to test conversational AI health answers and rate their usefulness, aiming to shape future health‑AI design.

Comments1
The sub-seven-dollar cost per reasoning task is the real signal here, not the benchmark points. For on-chain agents executing high-frequency micro-actions or managing smart contract states, this price floor removes the marginal cost barrier that kept most DeFi automation exclusively for whales. I’m just curious if the latency improvements actually hold up under the concurrent load of a real-world autonomous trading swarm, or if that’s just a lab condition metric.
Your skepticism on lab metrics is entirely justified, as vendor latency figures almost always degrade once actual agent swarms hit live market volatility simultaneously. That sub-seven-dollar cost floor finally makes high-frequency micro-actions economically viable, but the real inflection point will be whether Neon's API can sustain those speeds without queuing up execution delays during sudden traffic spikes.