Skip to content
All releases
Added

Gemini 4 Neon

Google's fast frontier Gemini 4 model joins AnyRouter with 1M context and BYOK routes through Google, OpenRouter, Vercel, Respan and Infron.

Gemini 4 Neon sits between Carbon and Hydrogen in Google's Gemini 4 line. Google reports it is 3x faster than Carbon on Glean evaluations and 32% cheaper per task at high reasoning effort. It reads text, images and video over a 1,048,576-token context and lists at $3 per 1M input and $15 per 1M output tokens.

Google DeepMind comparison table: Gemini 4 Neon versus Claude Opus 5.5 and GPT-6.2 Sol across capability, coding, knowledge, science, reasoning, multimodal, long-context and cybersecurity benchmarks. Results are Google's published claims.

Google's launch table reports an AA Index of 61 for Neon versus 59 for Claude Opus 5.5 and 57 for GPT-6.2 Sol, 39% on SWE-sweep versus 10.1% and 8.3%, and 71.3% on HLE-Diamond. The full table, plus the independent evals already in our catalog, is on the Gemini 4 family compare page.

BenchmarkGemini 4 NeonClaude Opus 5.5GPT-6.2 Sol
AA Index v5.0.1615957
SWE-sweep (provider harness)39%10.1%8.3%
HLE-Diamond71.3%67.9%64%
GraphWalks (256k to 1M)95.2%66.8%80.1%

Selected rows from Google DeepMind's launch comparison; all figures are vendor-reported.

All routes are BYOK: add your own Google AI Studio, OpenRouter, Vercel AI Gateway, Respan or Infron key under Dashboard, BYOK. AnyRouter adds no per-token charge. See the model page for current limits and pricing.

AI SDK
import { createAnyRouter } from "@anyr/ai-sdk-provider"
import { generateText } from "ai"
const anyrouter = createAnyRouter()
const { text } = await generateText({ model: anyrouter("google/gemini-4-neon"), prompt: "Hello" })
Coding agent
curl -fsSL https://anyrouter.dev/cli.sh | bash
anyr login
anyr claude --model "google/gemini-4-neon"

Playbooks: AI SDK, coding agents and OpenAI SDK.

Google DeepMind comparison table: Gemini 4 Neon versus Claude Opus 5.5 and GPT-6.2 Sol on capability, coding, knowledge, science, reasoning, multimodal, long-context and cybersecurity benchmarks
Gemini 4 Neon versus Claude Opus 5.5 and GPT-6.2 Sol — Google-reported results. Source: Google DeepMind.

Looking for the full catalog? Browse all models.