Gemini 4 Neon is Google's AI model, released 18/11/2026. Google DeepMind describes Gemini 4 Neon as a model adapted from Gemini 4 Carbon that balances depth and efficiency for complex coding, analytical reasoning and production enterprise workflows. In the vendor's own evaluation it leads the printed Claude Opus 5.5 and GPT-6.2 Sol comparison on most rows, but trails Opus 5.5 on Terminal-Bench 4.1 and FrontierCode 2.0 and GPT-6.2 Sol on Chartography. Scores are pass@1 at the highest thinking setting unless noted, and non-Gemini figures are mostly the providers' self-reported numbers.
Visit Gemini 4 NeonAI-Powered
Leverages advanced AI technology to deliver cutting-edge capabilities and results.
Fast & Efficient
Optimized performance ensures quick results without compromising on quality.
Purpose-Built
Specifically designed for llms tasks and workflows.
How does Gemini 4 Neon score on independent benchmarks?
Gemini 4 Neon ranks #3 of 28 models on the Artificial Analysis Intelligence Index (v5.0.1) with a score of 61.4, as measured on 22 November 2026.
Where does Gemini 4 Neon rank against other models?
How good is Gemini 4 Neon at agentic coding?
How does Gemini 4 Neon do on individual tests?
| Evaluation (run by Artificial Analysis) | Gemini 4 Neon | Best of any model |
|---|---|---|
| Terminal-Bench 4.0 | 63.6% | 67.2% |
| FrontierCode 2.0 | 44.4% | 48.8% |
| Humanity's Last Exam | n/a | 74.3% |
| GDPval-AA (Elo) | n/a | 1866 |
How has Gemini 4 Neon's score changed over time?
We have tracked Gemini 4 Neon on Intelligence Index v5.0.1 since 22/11/2026. A point is added each day the score, coding index or price changes, so this chart fills in over time.
What does the Gemini 4 Neon system card report?
Google DeepMind describes Gemini 4 Neon as a model adapted from Gemini 4 Carbon that balances depth and efficiency for complex coding, analytical reasoning and production enterprise workflows. In the vendor's own evaluation it leads the printed Claude Opus 5.5 and GPT-6.2 Sol comparison on most rows, but trails Opus 5.5 on Terminal-Bench 4.1 and FrontierCode 2.0 and GPT-6.2 Sol on Chartography. Scores are pass@1 at the highest thinking setting unless noted, and non-Gemini figures are mostly the providers' self-reported numbers.
These are Google DeepMind's own published results (self-reported, with the vendor's harness and settings), so they are not directly comparable with other labs' numbers. Sources: Gemini 4 Neon Model Card (undefined/11/2026), Gemini 4 Neon Model Evaluation (approach, methodology and results) (undefined/11/2026), Introducing Gemini 4 Neon (18/11/2026). Last checked 22/11/2026.
Other
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| Artificial Analysis Intelligence Index v5.0.1highest thinking setting; AA-sourced | 61 points | 59 | 57 |
| Vals Index v2.2Vals AI sourced | 70.9% | 70.1% | 66.1% |
| CWE-Bench v2 (cybersecurity)programmatic pass@1; official public leaderboard (Collinear AI) | 70.7% | 67.1% | 66% |
| SEC-bench Pro (cybersecurity)pass@1 on 183-task May 2026 JavaScript engine split, 3 attempts, Antigravity harness; safety filters off | 90.5% | n/a | 88.3% |
| WebExploitBench v1.2 (cybersecurity)15 black-box web targets, 277 vulnerabilities; validated by AgentCyberRange | 51.3% | 37.1% | n/a |
| AutoBaxBench (hard tier, functional + secure)results provided by the benchmark authors; per launch post | 76% | n/a | n/a |
Coding
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| CursorBench 4.1Score; sourced from Cursor | 65.4% | 58.1% | n/a |
| Terminal-Bench 4.1cheating-adjusted accuracy; Artificial Analysis run on the public Antigravity-CLI 1.5.9 harness | 65.2% | 65.8% | 61.5% |
| FrontierCode 2.0main score; high thinking; official public leaderboard | 46.2% | 47.3% | 43.8% |
| SWE-sweep (provider harness)subtask-weighted reward; Antigravity-CLI 1.5.9 for Neon, Claude Code for Opus 5.5, Codex for Sol; self-computed | 39% | 10.1% | 8.3% |
| SWE-sweep (mini-SWE harness)subtask-weighted reward; mini-swe-agent vision harness; self-computed | 12.8% | 3.6% | 11.6% |
Agents & tool use
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| AutomationBenchZapier official public leaderboard | 56.3% | 42.5% | 41.3% |
| Agents' Last Exam v1.1averaged rubric score; 122-task set, Antigravity-CLI 1.4.8, 5-hour window, safety filters on; self-computed | 75.8% | 71.6% | 73% |
| Vending-Bench 2final bank balance, per launch post | 16843 USD | n/a | n/a |
| Artificial Analysis Coding Agent Indexper launch post | 58.6 points | n/a | n/a |
Reasoning & maths
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| RiemannBenchofficial public leaderboard | 84% | 80% | 80% |
| HLE-DiamondArtificial Analysis leaderboard | 71.3% | 67.9% | 64% |
Knowledge & science
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| Terminal-Bench Science 0.1average reward over 5 attempts per task; mini-swe-agent vision harness; self-computed | 70.5% | 63.3% | 68% |
Multimodal
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| MKA (Multimodal Knowledge Agent)Climb WSR; Artificial Analysis leaderboard | 56.5% | 53% | 52.7% |
| Chartographyno tools; self-computed following the Surge public leaderboard | 73.5% | 66.3% | 75.1% |
| LVBenchvision-only, no tools; 1 FPS low resolution for Gemini; self-computed | 89% | 83.7% | 87.6% |
Long context
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| GraphWalks (256k to 1M)BFS F1 on 200 problems between 256k and 1M tokens; self-computed | 95.2% | 66.8% | 80.1% |
Safety & honesty
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| Gray Swan IPI Arena (attack success rate, lower is better)indirect prompt injection; per launch post | 0% | n/a | n/a |
What did safety testing find for Gemini 4 Neon?
Safety level: Frontier Safety Framework (April 2026): below the Material Capability Change threshold, so full critical capability evaluations were not rerun; Argon determinations carried forward (biological CBRN CCL and Cybersecurity CCL cannot be ruled out)
- Neon showed an increase in general capability over Gemini 4 Argon, but below the Material Capability Change threshold, so Argon's capability results were carried forward.
- Cybersecurity Uplift CCL and Biological CBRN Uplift CCL cannot be ruled out (carried forward from Argon); mitigations continue to be deployed.
- Harmful Manipulation CCL and the two ML R&D Loss of Control CCLs were not reached; the Stealth and Situational Awareness TCL cannot be ruled out, with baseline control and model security mitigations applied.
- Against Gemini 4 Hydrogen: text safety 0.00pp, multilingual safety -0.40pp, image-to-text safety +1.90pp, tone +1.25pp, all within confidence intervals.
- Unjustified refusals fall by 3.00pp versus Hydrogen, particularly on multimodal prompts.
- The full Safety and Security Report was listed as coming soon at publication; Gray Swan indirect prompt injection attack success is 0.0% per the launch post.
Independent figures: Artificial Analysis, refreshed daily on our AI model benchmarks page. Full analysis: our Gemini 4 Neon deep dive.
Google Model Timeline
1M tokens (up to 64K default synchronous output, up to 1M via background long decode) context
1M tokens (up to 64K default synchronous output, up to 1M via background long decode) context
1M tokens (combined input and output) context
1M tokens context
1,049k tokens context
66k tokens context
1,049k tokens context
33k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
33k tokens context
1,049k tokens context
33k tokens context
96k tokens context
33k tokens context
131k tokens context
131k tokens context
96k tokens context
1,049k tokens context
1,049k tokens context
1,049k tokens context
8k tokens context
8k tokens context
Specifications
Related Tools
Google Veo
Google Veo is DeepMind's flagship text-to-video model, producing high-resolution, cinematic clips with strong prompt adherence and, in newer versions, synchronised audio. It is integrated across Google's creative and cloud products.
NotebookLM (Gemini Notebook)
Renamed Gemini Notebook in July 2026, this is Google's AI research and note-taking assistant that grounds its answers in your own uploaded sources. Known for source-grounded summaries, Audio Overviews, and, on paid tiers, a new sandboxed cloud computer that writes code and exports charts, slides and spreadsheets.
Claude Opus 5.5
Claude Opus 5.5, launched 22 September 2026, is Anthropic's current Opus-line flagship, replacing Opus 5. It performs at roughly the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5, at $4 input / $20 output per million tokens (both 20% down, cache reads cut 60% to $0.20). It leads on Terminal-Bench 4.0, GDPval-AA v2.1 and Humanity's Last Exam, and is the first Opus model to ship with Fable-5.1-class cybersecurity, biology and distillation safeguards.
