Gemini 4 Neon Benchmarks: Did Google Really Beat Claude?
Google has officially announced Gemini 4 Neon, its latest model, which competes with Claude Opus 5.5 and the recently announced GPT-6.2 Sol. On the benchmarks Google shared, Neon beats both Claude Opus 5.5 and GPT-6.2 Sol on most tests, some by a huge margin. In this post, we compare the new Neon model with the other two across plenty of AI benchmarks.
All benchmarks
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| AA Index v5.0.1 | 61 | 59 | 57 |
| Vals Index v2.2 | 70.9% | 70.1% | 66.1% |
| CursorBench 4.1 | 65.4% | 58.1% | - |
| Terminal-Bench 4.1 | 65.2% | 65.8% | 61.5% |
| FrontierCode 2.0 | 46.2% | 47.3% | 43.8% |
| SWE-sweep (provider harness) | 39.0% | 10.1% | 8.3% |
| SWE-sweep (Mini-SWE harness) | 12.8% | 3.6% | 11.6% |
| AutomationBench | 56.3% | 42.5% | 41.3% |
| Agents' Last Exam v1.1 | 75.8% | 71.6% | 73.0% |
| RiemannBench | 84.0% | 80.0% | 80.0% |
| Terminal-Bench Science 0.1 | 70.5% | 63.3% | 68.0% |
| HLE-Diamond | 71.3% | 67.9% | 64.0% |
| MKA (Climb WSR) | 56.5% | 53.0% | 52.7% |
| Chartography | 73.5% | 66.3% | 75.1% |
| LVBench | 89.0% | 83.7% | 87.6% |
| GraphWalks (256K to 1M) | 95.2% | 66.8% | 80.1% |
| CWE-Bench v2 | 70.7% | 67.1% | 66.0% |
| SEC-bench Pro | 90.5% | - | 88.3% |
| WebExploitBench v1.2 | 51.3% | 37.1% | - |
A dash means no score has been reported for that model.
Winner by category
| Category | Winner |
|---|---|
| Overall intelligence: AA Index | Gemini 4 Neon |
| Professional work: Vals Index | Gemini 4 Neon |
| CursorBench | Gemini 4 Neon |
| Terminal-Bench 4.1 | Claude Opus 5.5 |
| FrontierCode 2.0 | Claude Opus 5.5 |
| SWE-sweep | Gemini 4 Neon |
| Automation | Gemini 4 Neon |
| Professional agents | Gemini 4 Neon |
| Mathematics: RiemannBench | Gemini 4 Neon |
| Scientific agent work | Gemini 4 Neon |
| Hard reasoning: HLE-Diamond | Gemini 4 Neon |
| Multimodal agent work | Gemini 4 Neon |
| Chart reasoning | GPT-6.2 Sol |
| Long video understanding | Gemini 4 Neon |
| Long context reasoning | Gemini 4 Neon |
| Vulnerability patching | Gemini 4 Neon |
| Vulnerability discovery | Gemini 4 Neon |
| Web exploitation | Gemini 4 Neon |
Leaderboard positions
| Benchmark | Gemini 4 Neon | Ranking |
|---|---|---|
| Artificial Analysis Intelligence Index v5.0.1 | 61 | #3 overall |
| Artificial Analysis Coding Agent Index | 58.6 | #1 |
| Vals Index v2.2 | 70.9% | #2 |
| Agent Arena Index | 1336 | #2 |
| AutomationBench | 56.3% | #1 |
| Vending-Bench 2 | $16,843 | #1 |
| Terminal-Bench Science 0.1 | 70.5% | Best in class |
| RiemannBench | 84.0% | Best in class |
| HLE-Diamond | 71.3% | Best in class |
Coding benchmarks
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| CursorBench 4.1 | 65.4% | 58.1% | - |
| Terminal-Bench 4.1 | 65.2% | 65.8% | 61.5% |
| FrontierCode 2.0 | 46.2% | 47.3% | 43.8% |
| SWE-sweep (provider harness) | 39.0% | 10.1% | 8.3% |
| SWE-sweep (Mini-SWE harness) | 12.8% | 3.6% | 11.6% |
Science and reasoning
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| Terminal-Bench Science 0.1 | 70.5% | 63.3% | 68.0% |
| RiemannBench | 84.0% | 80.0% | 80.0% |
| HLE-Diamond | 71.3% | 67.9% | 64.0% |
| Agents' Last Exam v1.1 | 75.8% | 71.6% | 73.0% |
Multimodal and long context
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| GraphWalks (256K to 1M) | 95.2% | 66.8% | 80.1% |
| LVBench | 89.0% | 83.7% | 87.6% |
| MKA (Climb WSR) | 56.5% | 53.0% | 52.7% |
| Chartography | 73.5% | 66.3% | 75.1% |
Cybersecurity
| Benchmark | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol |
|---|---|---|---|
| SEC-bench Pro | 90.5% | - | 88.3% |
| CWE-Bench v2 | 70.7% | 67.1% | 66.0% |
| WebExploitBench v1.2 | 51.3% | 37.1% | - |
| AutoBaxBench (hard tier) | 76% | - | - |
API pricing
| Model | Input, per 1M | Cached input, per 1M | Output, per 1M |
|---|---|---|---|
| Gemini 4 Neon | $3.00 | $0.15 | $15.00 |
| GPT-6.2 Sol | $2.00 | $0.10 | $10.00 |
| Claude Opus 5.5 | $4.00 | $0.20 | $20.00 |
Thinking levels
| Setting | Gemini 4 Neon |
|---|---|
| Low | Yes |
| Medium | Yes |
| High | Yes, default |
| xhigh | Yes, highest |
Specifications
| Specification | Gemini 4 Neon |
|---|---|
| Model ID | gemini-4-neon |
| Model family | Gemini 4 |
| Base model | Adapted from Gemini 4 Carbon |
| Context window | 1M tokens, input and output combined |
| Default maximum output | 64K tokens |
| Background long decode | Up to 1M output tokens |
| Context in Antigravity | Up to 512K tokens |
| Knowledge cutoff | March 2026 |
| Text input | Yes |
| Image input | Yes |
| Audio input | Yes |
| Video input | Yes |
| Output | Text |
| Reasoning | Yes |
| Default thinking | High |
| Long-running agents | Yes |
| Coding agents | Yes |
| Multimodal knowledge work | Yes |
| Cybersecurity workflows | Yes |
| Release date | November 18, 2026 |
Availability
| Platform | Gemini 4 Neon |
|---|---|
| Gemini API | Paid tier |
| Google AI Studio | Yes |
| Gemini app | Google AI Pro and Ultra |
| Gemini for business | Yes |
| Google Antigravity | Paid Google AI Pro and Ultra |
| Orchestrator in Antigravity | Yes |
| Android Studio | Yes |
| Free Gemini app users | No direct model selection |
| Google AI Plus | Not included in Antigravity |
| Fairwind Program (cyber defenders) | Rolling out to eligible defenders |
Head-to-head summary
| Category | Gemini 4 Neon | Claude Opus 5.5 | GPT-6.2 Sol | Winner |
|---|---|---|---|---|
| Overall intelligence | 61 | 59 | 57 | Neon |
| Knowledge work | 70.9% | 70.1% | 66.1% | Neon |
| Cursor coding | 65.4% | 58.1% | - | Neon |
| Terminal coding | 65.2% | 65.8% | 61.5% | Opus 5.5 |
| Frontier coding | 46.2% | 47.3% | 43.8% | Opus 5.5 |
| SWE-sweep | 39.0% | 10.1% | 8.3% | Neon |
| Automation | 56.3% | 42.5% | 41.3% | Neon |
| Professional agents | 75.8% | 71.6% | 73.0% | Neon |
| Science | 70.5% | 63.3% | 68.0% | Neon |
| Hard reasoning | 71.3% | 67.9% | 64.0% | Neon |
| Long context | 95.2% | 66.8% | 80.1% | Neon |
| Chart reasoning | 73.5% | 66.3% | 75.1% | GPT-6.2 Sol |
| Video understanding | 89.0% | 83.7% | 87.6% | Neon |
| Cyber patching | 70.7% | 67.1% | 66.0% | Neon |
| API input price | $3 | $4 | $2 | GPT-6.2 Sol |
| API output price | $15 | $20 | $10 | GPT-6.2 Sol |