by Baidu
ERNIE-4.5-VL-424B-A47B is a multimodal Mixture-of-Experts (MoE) model from Baidu’s ERNIE 4.5 series, featuring 424B total parameters with 47B active per token. It is trained jointly on text and image data using a heterogeneous MoE architecture and modality-isolated routing to enable high-fidelity cross-modal reasoning, image understanding, and long-context generation (up to 131k tokens). Fine-tuned with techniques like SFT, DPO, UPO, and RLVR, this model supports both “thinking” and non-thinking inference modes. Designed for vision-language tasks in English and Chinese, it is optimized for efficient scaling and can operate under 4-bit/8-bit quantization.
| Signal | Strength | Weight | Impact |
|---|---|---|---|
| Recency2026-03-03T20:31:18.422Z | 88 | 15% | +13.2 |
| Context Window2026-03-03T20:31:18.422Z | 81 | 15% | +12.1 |
| Capabilities2026-03-03T20:31:18.422Z | 43 | 25% | +10.7 |
| Output Capacity2026-03-03T20:31:18.422Z | 70 | 10% | +7.0 |
| Versatility2026-03-03T20:31:18.422Z | 50 | 10% | +5.0 |
| Pricing Tier2026-03-03T20:31:18.422Z | 1 | 25% | +0.3 |
Cost Estimator
You save $34.10/month vs category average