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-oss-20b (free)
| Gemini 3.6 Flash | gpt-oss-20b (free) OpenAI | |
|---|---|---|
| Pricing | ||
| Input $/1M | $1.50 | Free |
| Output $/1M | $7.50 | Free |
| Blended $/1M (3:1) | $3.00 | Free |
| Cache read $/1M | $0.15 | — |
| Cache write $/1M | $0.083 | — |
| Quality indices | ||
| Intelligence Index | 50.1 | 14.9 |
| Coding Index | 69.2 | 20.7 |
| Math Index | — | 89.3 |
| Agentic Index | 38.7 | 3.1 |
| Speed & latency | ||
| Output speed (tok/s) | 311 | 243 |
| Time to first token | 12.80s | 0.39s |
| Time to first answer token | 12.80s | 8.61s |
| Benchmarks (Artificial Analysis) | ||
| MMLU-Pro | — | 74.8% |
| GPQA Diamond | 92.8% | 68.8% |
| Humanity's Last Exam | 38.3% | 9.8% |
| LiveCodeBench | — | 77.7% |
| SciCode | 52.7% | 34.4% |
| AIME 2025 | — | 89.3% |
| IFBench | — | 65.1% |
| AA-LCR (Long Context Reasoning) | 69.7% | 30.7% |
| Terminal-Bench Hard | — | 10.6% |
| Terminal-Bench 2.1 | 77.5% | 13.9% |
| τ²-Bench (Telecom) | — | 60.2% |
| τ³-Bench Banking | 24.5% | 7.0% |
| Design Arena (Elo) | ||
| Website | 1,332 | 880 |
| Data viz | — | 971 |
| Specs | ||
| Context window | 1.05M | 131K |
| Max output tokens | 66K | 33K |
| Input modalities | text, image, video, file, audio | text |
| Tokenizer | Gemini | GPT |