Google's first 1M-context model — multimodal needle-in-haystack champion
Google's next-generation model with breakthrough long context capabilities
| Benchmark | Gemini 1.5 Pro | Gemini Pro | Δ |
|---|---|---|---|
| MMLU | 81.9% | 79.1% | +2.8 |
| HumanEval | 71.9% | 67.7% | +4.2 |
| MATH | 58.5% | 32.6% | +25.9 |