Baidu: ERNIE 4.5 VL 424B A47B

Description

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.

Raw fields snapshot

{
  "id": "baidu/ernie-4.5-vl-424b-a47b",
  "name": "Baidu: ERNIE 4.5 VL 424B A47B ",
  "description": "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.",
  "created": 1751300903,
  "canonical_slug": "baidu/ernie-4.5-vl-424b-a47b",
  "hugging_face_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
  "source_type": "openrouter_only",
  "context_length": 123000,
  "max_completion_tokens": 16000,
  "is_moderated": false,
  "architecture": {
    "modality": "text+image->text",
    "input_modalities": [
      "image",
      "text"
    ],
    "output_modalities": [
      "text"
    ],
    "tokenizer": "Other",
    "instruct_type": null
  },
  "input_modalities": [
    "image",
    "text"
  ],
  "output_modalities": [
    "text"
  ],
  "modality": "text+image->text",
  "tokenizer": "Other",
  "instruct_type": null,
  "supported_parameters": [
    "frequency_penalty",
    "include_reasoning",
    "max_tokens",
    "presence_penalty",
    "reasoning",
    "repetition_penalty",
    "seed",
    "stop",
    "temperature",
    "top_k",
    "top_p"
  ],
  "default_parameters": {},
  "per_request_limits": null,
  "top_provider": {
    "context_length": 123000,
    "max_completion_tokens": 16000,
    "is_moderated": false
  },
  "pricing": {
    "prompt": "0.00000042",
    "completion": "0.00000125"
  },
  "PPM": {
    "prompt": 0.42,
    "completion": 1.25
  },
  "openrouter_raw": {
    "id": "baidu/ernie-4.5-vl-424b-a47b",
    "canonical_slug": "baidu/ernie-4.5-vl-424b-a47b",
    "hugging_face_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
    "name": "Baidu: ERNIE 4.5 VL 424B A47B ",
    "created": 1751300903,
    "description": "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.",
    "context_length": 123000,
    "architecture": {
      "modality": "text+image->text",
      "input_modalities": [
        "image",
        "text"
      ],
      "output_modalities": [
        "text"
      ],
      "tokenizer": "Other",
      "instruct_type": null
    },
    "pricing": {
      "prompt": "0.00000042",
      "completion": "0.00000125"
    },
    "top_provider": {
      "context_length": 123000,
      "max_completion_tokens": 16000,
      "is_moderated": false
    },
    "per_request_limits": null,
    "supported_parameters": [
      "frequency_penalty",
      "include_reasoning",
      "max_tokens",
      "presence_penalty",
      "reasoning",
      "repetition_penalty",
      "seed",
      "stop",
      "temperature",
      "top_k",
      "top_p"
    ],
    "default_parameters": {},
    "expiration_date": null
  }
}