Day 7 receipt
Gemini 4 Neon
Day 4 of 7Claims on the record
We check on 25 Nov 2026
- They promised
- Frontier-level coding and analysis at faster speed and lower cost per task than Carbon, available to Pro and Ultra subscribers and in the API.
The claims we’ll check
- “Neon ranks ahead of comparable models and #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level”
- “ranking #1 overall on the”
- “a 32% lower cost per task than Carbon”
Quoted word for word from Introducing Gemini 4 Neon. Day 1: independent testing confirms a third-place score and faster runs, but the lead on coding agents is one point on Google's own tool, and the cost saving is smaller than Google states.
Full evidence 3 claims · 0 user reports · 3 sources
Google's fourth Gemini 4 model in about a month, positioned between Carbon and Hydrogen for coding, knowledge work and science. Live in the API, Antigravity and the Gemini app for paid Pro and Ultra subscribers.
Launch post: Introducing Gemini 4 Neon
Why this stamp
Day 1: independent testing confirms a third-place score and faster runs, but the lead on coding agents is one point on Google's own tool, and the cost saving is smaller than Google states.
Google says Neon ranks third overall on the Artificial Analysis index and first on its coding-agent index, at a 32% lower cost per task than Carbon. Artificial Analysis confirms the third place, a 1-point gain over Carbon, and about 40% less time per task. It puts the coding-agent lead at one point, using Google's own Antigravity CLI, and the cost saving at about 17% at the top thinking level. SiliconANGLE notes Google's comparisons are its own tests and that some buyers may not find the premium over cheaper rivals worth it. Several benchmark leads in the launch post come from third-party leaderboards Google cites but we have not rerun. No hands-on user reports yet. We re-stamp on day 7 (25 November) and day 14.
Launch dayWhat they claimed, quoted word for word · 3
“Neon ranks ahead of comparable models and #3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level”
Artificial Analysis confirms 61 and third place, behind GPT-6.2 Astra and Claude Fable 5.2, and calls it only a 1-point gain over Gemini 4 Carbon.
“ranking #1 overall on the”
Artificial Analysis shows Neon at 59 on its coding-agent index with Google's own Antigravity CLI, against 58 for Claude Code with Fable 5.2 and with Opus 5.5. A one-point lead on a harness Google makes; not yet reproduced elsewhere.
“a 32% lower cost per task than Carbon”
Artificial Analysis measures about 17% lower cost per task at its top thinking level, and cost parity at half the price at high effort only against Carbon's standard rates. Neon also uses about 15% more output tokens per task.
OthersWhat outlets and testers say · 3
Independent index: 61, third overall, a small gain over Carbon with about 40% less time per task; coding-agent score in line with Claude Code on Fable 5.2 and Opus 5.5.
“Google continues to accelerate its release cadence with Gemini 4 Neon” · Open the source
Notes the benchmarks are Google's own, repeats the Artificial Analysis ranking, and questions whether Neon's premium over cheaper rivals is justified.
“Google’s latest midtier model Gemini 4 Neon is three times faster than Carbon and wins on some metrics” · Open the source
Confirms the same-day API availability and the 1M-token context, and places Neon as Google's counterpart to Claude Opus and GPT Sol.
“Google ships Gemini 4 Neon, its fourth new Gemini model in a month” · Open the source