Gemini 4 Neon vs GPT-6.2 Sol: Code Red for OpenAI?
Google has officially announced Gemini 4 Neon, just a day after OpenAI launched GPT-6.2 Sol. On Google's benchmarks, Neon beats GPT-6.2 Sol on 16 of 17 tests, and it leads the Artificial Analysis Intelligence Index 61.4 to 57.3. But Sol is still far cheaper to run, at about 84% less per task, and it finishes tasks in less than half the time. 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 | 57 | |
| Vals Index v2.2 | 70.9% | 66.1% | |
| Terminal-Bench 4.1 | 65.2% | 61.5% | |
| FrontierCode 2.0 | 46.2% | 43.8% | |
| SWE-sweep (provider harness) | 39.0% | 8.3% | |
| SWE-sweep (Mini-SWE harness) | 12.8% | 11.6% | |
| AutomationBench | 56.3% | 41.3% | |
| Agents' Last Exam v1.1 | 75.8% | 73.0% | |
| RiemannBench | 84.0% | 80.0% | |
| Terminal-Bench Science 0.1 | 70.5% | 68.0% | |
| HLE-Diamond | 71.3% | 64.0% | |
| MKA (Climb WSR) | 56.5% | 52.7% | |
| Chartography | 73.5% | 75.1% | |
| LVBench | 89.0% | 87.6% | |
| GraphWalks (256K to 1M) | 95.2% | 80.1% | |
| CWE-Bench v2 | 70.7% | 66.0% | |
| SEC-bench Pro | 90.5% | 88.3% | |
Artificial Analysis benchmarks at the highest effort | |||
| Intelligence Index v5.0.1 | 61.4 | 57.3 | |
| AA-Briefcase v1.1 | 1700 Elo | 1612 Elo | |
| AutomationBench-AA | 79.3% | 71.3% | |
| Terminal-Bench 4.0 | 63.6% | 60.1% | |
| FrontierCode-AA 2.0 | 44.4% | 44.4% | |
| AA-Omniscience v1.1 | 54.8 | 40.4 | |
| GDP.pdf | 28.0% | 30.0% | |
| HLE-Diamond | 71.3% | 64.0% | |
| CritPt v2 (Hard Subset) | 60.5% | 62.5% | |
| Terminal-Bench Science 0.1 | 70.0% | 67.6% | |
Speed, cost and tokens at the highest effort | |||
| Cost per Intelligence Index task | $6.44 | $1.01 | |
| Time per task | 21.5 min | 8.5 min | |
| Output tokens across the Index | 283M | 60.9M | |
| Input tokens across the Index | 8.12B | 2.22B | |
| Output speed | 110 tokens/s | 79 tokens/s | |
| Time to first answer token | 56.5 s | 75.1 s | |
Pricing and specs | |||
| Input, per 1M tokens | $3.00 | $2.00 | |
| Output, per 1M tokens | $15.00 | $10.00 | |
| Cached input, per 1M tokens | $0.15 | $0.10 | |
| Cache writes, per 1M tokens | Implicit caching only | $2.50 for 30 min | |
| Batch input, per 1M tokens | $1.50 | $1.00 | |
| Batch output, per 1M tokens | $7.50 | $5.00 | |
| Context window | 1.05M tokens | 1.05M tokens | |
| Standard max output | 64K tokens | 128K tokens | |
| Extended output | Up to 1M with long decode | 128K tokens | |
| Reasoning levels | Low, medium, high and xhigh | Low to max | |
| 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 1, 2026 | |
| Released | November 18, 2026 | November 17, 2026 | |