Gemini 4 Neon
Google's new mid-tier Gemini 4 model — frontier-level reasoning for complex coding, knowledge work and science at Flash speeds, $3/$15 per 1M tokens with a 1M-token context window.
Why it matters
Ranks #3 overall (score 61) on the Artificial Analysis Intelligence Index at xhigh thinking, #1 on the Coding Agent Index, and Google claims 32% lower cost per task than Carbon — near-frontier quality at a repeatable daily-driver price.
What you could build with it
A software team could route its coding-agent fleet through this model as the default planner, reserving the heaviest frontier model only for tasks it fails, and cut agent-task spend by roughly a third while keeping near-frontier success rates on multi-file refactors.
Does it hold up?
Too early to judge — only day-0 independent testing exists so far; the GeminiPulse gauntlet shows strong single-shot coding and fast security auditing at High thinking, but no production usage has been reported yet.
Built with Gemini 4 Neon
- We Tested Gemini 4 Neon Against Its Siblings & Claude Opus 5.5geminipulse.com · Day-0 independent gauntlet: Neon (High) scored 100/100 on single-shot frontend builds in 421s, and Neon (Low) caught all 6 adversarial security traps in 50s.
- Gemini 4 Neon Just Beat Carbon (And Pro Users Get It)youtube.com · Video breakdown of the launch, pricing and the Dec 1 price change, scoring Neon top on 16 of 19 benchmark rows in Google's own table.
- Gemini 4 Neon launches: Carbon quality at Flash speeds, for AI Pro9to5google.com · Launch-day report confirming Neon replaces Argon in the Gemini app for Pro/Ultra subscribers and its $3/$15 pricing and 1M context.
Learn more
First spotted on article: source.
More AI models
Edge0-145B: open 145B MoE that runs on a single consumer GPUEdge0 open-sourced 145B and 120B latent-MoE models derived from Qwen3.8-Flash-Next, plus an inference framework that…models · JEV 0.74Strata: run a 125B model on a gaming PCFree, MIT-licensed local inference engine that runs Qwen3.8-Flash-Next (125B MoE, about 6B active per token) on…models · JEV 0.73Qwen4-Max: Alibaba's new flagship model goes liveAlibaba's 2.4T-parameter flagship is now generally available via API, with a 1M-token context, native vision, and…models · JEV 0.67smolmaxxing: run 16B LLMs on a 1 GB TV box or a 2013 iPadAn MIT-licensed llama.cpp patch set that runs language models up to a 16B MoE on devices with about 1 GB of usable RAM…models · JEV 0.65Blue TTS v3: a 388MB open text-to-speech model with voice cloningAn open-source 388MB text-to-speech model (MIT) with voice cloning from about 10 seconds of speech, 7 languages, 15…models · JEV 0.65Haru-4BOpen full-duplex spoken-dialogue foundation model that listens and speaks simultaneously (7.5 Hz frames, 133 ms per…models · JEV 0.63
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.