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-5.2
| Gemini 3.6 Flash | GPT-5.2 OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $1.50 | $1.75 |
| Output $/1M | $7.50 | $14.00 |
| Blended $/1M (3:1) | $3.00 | $4.81 |
| Cache read $/1M | $0.15 | $0.175 |
| Cache write $/1M | $0.083 | — |
| Quality indices | ||
| Intelligence Index | 50.1 | 42.2 |
| Coding Index | 69.2 | — |
| Math Index | — | 99 |
| Agentic Index | 38.7 | — |
| Speed & latency | ||
| Output speed (tok/s) | 311 | 85 |
| Time to first token | 12.80s | 111.04s |
| Time to first answer token | 12.80s | 111.04s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 87.4% |
| GPQA Diamond | 92.8% | 90.3% |
| Humanity's Last Exam | 38.3% | 35.4% |
| LiveCodeBench | — | 88.9% |
| SciCode | 52.7% | 52.1% |
| AIME 2025 | — | 99.0% |
| IFBench | — | 75.4% |
| AA-LCR (Long Context Reasoning) | 69.7% | 72.7% |
| Terminal-Bench Hard | — | 47.0% |
| Terminal-Bench 2.1 | 77.5% | — |
| τ²-Bench (Telecom) | — | 84.8% |
| τ³-Bench Banking | 24.5% | — |
| Design Arena (Elo) | ||
| Website | 1,332 | 1,222 |
| Web apps | — | 1,142 |
| Full-stack | — | 1,100 |
| Mobile apps | — | 1,167 |
| Android native | — | 1,052 |
| UI components | — | 1,231 |
| Data viz | — | 1,238 |
| SVG | — | 1,185 |
| 3D | — | 1,138 |
| Game dev | — | 1,253 |
| Godot game dev | — | 1,142 |
| Code categories | — | 1,203 |
| ASCII art | — | 1,194 |
| Specs | ||
| Context window | 1.05M | 400K |
| Max output tokens | 66K | 128K |
| Input modalities | text, image, video, file, audio | file, image, text |
| Tokenizer | Gemini | GPT |