Gemini 4 Neon vs Claude Opus 5.5: Neon Beats Opus 5.5?
Google has officially announced Gemini 4 Neon, and on Google's own benchmarks it beats Claude Opus 5.5 on most tests. On the Artificial Analysis Intelligence Index, Neon scores 61.4 at xhigh against 59.1 for Opus 5.5 at max, and it costs about 36% less per task. Opus still wins a few, including Terminal-Bench 4.1, FrontierCode and AA-Briefcase. Here are the detailed benchmarks to help you see the real differences between these models.
Google's benchmarks | |||
| Benchmark | Difference | ||
|---|---|---|---|
| AA Intelligence Index v5.0.1 | 61 | 59 | |
| Vals Index v2.2 | 70.9% | 70.1% | |
| CursorBench 4.1 | 65.4% | 58.1% | |
| Terminal-Bench 4.1 | 65.2% | 65.8% | |
| FrontierCode 2.0 | 46.2% | 47.3% | |
| SWE-sweep (provider harness) | 39.0% | 10.1% | |
| SWE-sweep (Mini-SWE harness) | 12.8% | 3.6% | |
| AutomationBench | 56.3% | 42.5% | |
| Agents' Last Exam v1.1 | 75.8% | 71.6% | |
| RiemannBench | 84.0% | 80.0% | |
| Terminal-Bench Science 0.1 | 70.5% | 63.3% | |
| HLE-Diamond | 71.3% | 67.9% | |
| MKA (Climb WSR) | 56.5% | 53.0% | |
| Chartography | 73.5% | 66.3% | |
| LVBench | 89.0% | 83.7% | |
| GraphWalks (256K to 1M) | 95.2% | 66.8% | |
| CWE-Bench v2 | 70.7% | 67.1% | |
| WebExploitBench v1.2 | 51.3% | 37.1% | |
| SEC-bench Pro | 90.5% | Not reported | |
Artificial Analysis benchmarks at the highest effort | |||
| Intelligence Index v5.0.1 | 61.4 | 59.1 | |
| AA-Briefcase v1.1 | 1700 Elo | 1800 Elo | |
| AutomationBench-AA | 79.3% | 69.5% | |
| Terminal-Bench 4.0 | 63.6% | 59.6% | |
| FrontierCode-AA 2.0 | 44.4% | 46.4% | |
| AA-Omniscience v1.1 | 54.8 | 48.1 | |
| GDP.pdf | 28.0% | 26.2% | |
| HLE-Diamond | 71.3% | 67.9% | |
| CritPt v2 (Hard Subset) | 60.5% | 68.0% | |
| Terminal-Bench Science 0.1 | 70.0% | 59.0% | |
Intelligence Index by reasoning effort | |||
| Max | No max level | 59 | |
| Xhigh | 61 | 57 | |
| High | 60 | 52 | |
| Medium | 55 | 48 | |
| Low | 44 | 35 | |
Cost per Intelligence Index task by reasoning effort | |||
| Max | No max level | $10.04 | |
| Xhigh | $6.44 | $6.05 | |
| High | $3.74 | $3.46 | |
| Medium | $1.90 | $2.45 | |
| Low | $0.75 | $0.98 | |
Speed, cost and tokens at the highest effort | |||
| Cost per Intelligence Index task | $6.44 | $10.04 | |
| Time per task | 21.5 min | 24.5 min | |
| Output tokens across the Index | 283M | 218M | |
| Output speed | 110 tokens/s | 118 tokens/s | |
| Time to first answer token | 56.5 s | 336 s | |
Pricing and specs | |||
| Input, per 1M tokens | $3.00 | $4.00 | |
| Output, per 1M tokens | $15.00 | $20.00 | |
| Cached input, per 1M tokens | $0.15 | $0.20 | |
| Batch input, per 1M tokens | $1.50 | $2.00 | |
| Batch output, per 1M tokens | $7.50 | $10.00 | |
| Context window | 1.05M tokens, input and output combined | 1M tokens | |
| Standard max output | 64K tokens | 128K tokens | |
| Extended output | Up to 1M with long decode | 300K in batch, beta | |
| Reasoning levels | Low, medium, high and xhigh | Low to max with adaptive thinking | |
| Text and image input | Yes | Yes | |
| Audio input | Yes | No | |
| Video input | Yes | No | |
| PDF input | Yes | Yes | |
| Computer use | Yes | Yes | |
| Code execution | Yes | Yes | |
| Free API tier | No | No | |
| Knowledge cutoff | March 2026 | June 2026 | |
| Released | November 18, 2026 | September 22, 2026 | |