In case you opened the Gemini app this morning looking for Gemini 4 Argon then you probably noticed it vanished overnight. Google quietly pulled Argon from the model picker and retired Gemini 3.1 Pro. It was swaped in a brand new mid tier release called Gemini 4 Neon.
Gemini 4 Neon was launched on November 18th 2026. And it is designed to bridge the gap between lightweight daily chats and heavy-duty research.
Google claims it delivers flagship level coding and analytical muscle at three times the speed of Gemini 4 Carbon along with a pricing twist you should know before switching your API keys.
Read more about the special features and benefits of Gemini 4 Neon which makes it better for Argon users hereon.
Where does Gemini 4 Neon Launch Fits in Google’s AI Lineup?
Google is naming its fourth generation models after elements on the periodic table which gets confusing fast.
Here are the models that we are taking about, which are named after natural elements and are of the active tiers:
Related Stories
-
Gemini 4 Hydrogen: It was launched November 12 for quick, everyday execution and low-cost sub-agent tasks.
Gemini 4 Neon: The new middle workhorse built for multi-step coding, spreadsheet analysis, and scientific research.
Gemini 4 Carbon: This one was launched on November 10.The heavy-duty flagship is built for massive, long-horizon engineering projects.
Neon is adapted directly from Carbon and shares its 1 million-token context window. You can feed it text, code, audio, video, PDFs, and dense spreadsheets in a single prompt. Developers can also dial its reasoning effort across four thinking levels: low, medium, high (default), and xhigh.
Why did Google Retire Gemini 4 Argon Overnight?
If you built workflows around Gemini 4 Argon (which only launched in late September in 2026) having it yanked from the Gemini app picker after seven weeks feels jarring.
Google hasn't published a formal post-mortem on Argon. But the benchmark math tells the story.
Neon beats Argon across agentic coding, tool use, and steerability. It is also capable of handling the exact same mid-tier workload.
Keeping four overlapping element tiers made little sense once Neon proved it could carry both fast execution and deep reasoning.
How Fast and Smart is Google’s Gemini 4 Neon?
Benchmark Performance
At its xhigh thinking setting, Neon scored 61 on the Artificial Analysis Intelligence Index v5.0.1. That puts it third overall.
It is trailing only Anthropic’s Claude Fable 5.2 (65) and OpenAI’s GPT-6.2 Astra (62) while edging past Carbon’s score of 60.
On practical work benchmarks, Neon posted:
-
58.6 (#1 at launch) on the Artificial Analysis Coding Agent Index
-
39.0% on SWE-sweep inside Google Antigravity
-
70.5% on Terminal-Bench Science and 84.0% on RiemannBench
-
#1 rankings on the Vals Legal Research Bench, Tax Agent Bench, and Zapier’s AutomationBench (56.3%)
The Speed and Hallucination Catch
Google highlights partner tests (such as enterprise search firm Glean). The report shows that Neon is running three times faster than Carbon at 60% of the cost.
Independent testing at default settings (high) tells a more grounded story: Neon finishes tasks in roughly half the time of Carbon and cuts per-task spend by about 32%.
There is a trade-off, too.
On AA-Omniscience v1.1, Neon recorded a 13.9% factual hallucination rate at high which is higher than Carbon (11.4%) and the retired Argon (4.3%). Though it is still far cleaner than rival models like Claude Opus 5.5 (50.3%).
What is Gemini 4 Neon Pricing and the November Catch
Neon costs $3.00 per million input tokens ($0.15 for cached inputs) and $15.00 per million output tokens for developers using the Gemini API.
Wait before migrating every production pipeline today, though.
Flagship Gemini 4 Carbon is currently running on introductory promotional pricing of $2.00 input / $10.00 output through November 30, 2026.
That makes Neon 50% more expensive per token than Carbon for the next eleven days. Once Carbon jumps to its standard $4.00 / $20.00 rate on December 1 Neon becomes the clear mid-tier bargain.
How to Access Gemini 4 Neon Today
If you pay for Google AI Pro or Ultra, Neon is already waiting in your Gemini app. Just pick "Deeper analysis" from the model dropdown. Free-tier users don't get access yet.
Developers can wire it into apps immediately through the paid Gemini API in Google AI Studio, Android Studio, and Gemini for Business.
If you use Google Antigravity, look for the new "Orchestrator (Neon)" mode. It lets Neon act as the lead architect planning big codebase changes while handing off grunt work to cheaper Hydrogen sub-agents.
One caveat for readers in India: users on telecom-bundled Pro plans (like the Jio offer) are reporting entitlement errors locking them out of Gemini 4 models inside Antigravity.
Security teams get a separate build. Google is rolling out Gemini 4 Neon for Cyber over the next few days, though access is restricted to vetted defenders inside Google Cloud’s Fairwind Program.
