Compare models
Pick up to 4 models for a side-by-side spec sheet. Hover any dotted metric name for what the test measures. Best value in each row is highlighted.
Gemini 3.6 FlashGPT-4o (batch)
| Gemini 3.6 Flash | GPT-4o (batch) OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $0.75 | $1.25 |
| Output $/1M | $3.75 | $5.00 |
| Blended $/1M (3:1) | $1.50 | $2.19 |
| Cache read $/1M | $0.075 | $0.625 |
| Cache write $/1M | $0.042 | — |
| Quality indices | ||
| Intelligence Index | 34 | 8.4 |
| Coding Index | 69.2 | — |
| Math Index | — | 6 |
| Agentic Index | 29 | — |
| Speed & latency | ||
| Output speed (tok/s) | 265 | 0 |
| Time to first token | 21.43s | 0s |
| Time to first answer token | 21.43s | 0s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 74.8% |
| GPQA Diamond | 92.8% | 54.3% |
| Humanity's Last Exam | 40.8% | 2.4% |
| LiveCodeBench | — | 30.9% |
| SciCode | 53.4% | — |
| MATH-500 | — | 75.9% |
| AIME | — | 15.0% |
| AIME 2025 | — | 6.0% |
| IFBench | — | 34.3% |
| AA-LCR (Long Context Reasoning) | 80.0% | 49.3% |
| Terminal-Bench Hard | — | 8.3% |
| Terminal-Bench 2.1 | 77.5% | — |
| τ²-Bench (Telecom) | — | 25.1% |
| τ³-Bench Banking | 29.9% | — |
| Design Arena (Elo) | ||
| Website | 1,312 | 843 |
| Web apps | 1,211 | — |
| Full-stack | 1,188 | — |
| Mobile apps | 1,228 | — |
| Android native | 1,213 | — |
| UI components | 1,318 | 901 |
| Data viz | 1,311 | 870 |
| 3D | 1,302 | 897 |
| Game dev | 1,286 | 930 |
| Agentic game dev | 1,180 | — |
| HTML slides | 1,152 | — |
| Python→PPTX slides | 1,146 | — |
| Code categories | 1,305 | 876 |
| ASCII art | 1,304 | — |
| Specs | ||
| Context window | 1.05M | 128K |
| Max output tokens | 66K | 16K |
| Input modalities | text, image, video, file, audio | text, image, file |
| Tokenizer | Gemini | GPT |