ERNIE 5.5 Flash Base vs ERNIE 5.5 Flash Preview vs Gemini Decision 1
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | |||
|---|---|---|---|
| Task | Base language modeling | Coding and language generation | Decision scoring |
| Published capabilities | Pretrained mixture-of-experts checkpoint with published FP8 and BF16 weights. | Post-trained from the Flash Base checkpoint for coding, design and knowledge work. | Evaluates up to 1,000 binary, multiple-choice or rubric questions over one shared input. |
| Model size | 280B total / 14B active | 280B total / 14B active | — |
| Access | Public weights for inference and fine-tuning | Baidu Qianfan API | Gemini Decisions API in Google AI Studio |
| Result | — | — | Probabilities for supplied answer options |
| Sources | Model weights | Publisher documentation | Publisher documentation |
| Inputs | Text | Text | Text, Images, Audio, Video, PDFs |
| Context window | 1,048,576 tokens | 1,048,576 tokens | 262,144 tokens |
| Input price per million tokens | — | — | $0.10 |
| Output price per million tokens | — | — | Free |
| Weights | Open weights | Proprietary | Proprietary |
| Released | Nov 5, 2026 | Nov 11, 2026 | Nov 11, 2026 |