Baidu: ERNIE 4.5 21B A3B
Server-rendered model summary page for indexing/share previews. Use the interactive explorer for full filtering and comparison.
Identifiers & provenance
- Primary ID
- baidu/ernie-4.5-21b-a3b
- OpenRouter ID
- baidu/ernie-4.5-21b-a3b
- Canonical slug
- baidu/ernie-4.5-21b-a3b
Source semantics
- Arena rank is a human-preference leaderboard signal, not a universal truth metric.
- OpenRouter usage/popularity reflects adoption/traffic, not benchmark quality.
- Pricing fields may differ by provider and can include extra modes beyond prompt/completion.
Read more on Methodology & data sources.
Description
A sophisticated text-based Mixture-of-Experts (MoE) model featuring 21B total parameters with 3B activated per token, delivering exceptional multimodal understanding and generation through heterogeneous MoE structures and modality-isolated routing. Supporting an extensive 131K token context length, the model achieves efficient inference via multi-expert parallel collaboration and quantization, while advanced post-training techniques including SFT, DPO, and UPO ensure optimized performance across diverse applications with specialized routing and balancing losses for superior task handling.
Raw fields snapshot
{
"id": "baidu/ernie-4.5-21b-a3b",
"name": "Baidu: ERNIE 4.5 21B A3B",
"description": "A sophisticated text-based Mixture-of-Experts (MoE) model featuring 21B total parameters with 3B activated per token, delivering exceptional multimodal understanding and generation through heterogeneous MoE structures and modality-isolated routing. Supporting an extensive 131K token context length, the model achieves efficient inference via multi-expert parallel collaboration and quantization, while advanced post-training techniques including SFT, DPO, and UPO ensure optimized performance across diverse applications with specialized routing and balancing losses for superior task handling.",
"created": 1755034167,
"canonical_slug": "baidu/ernie-4.5-21b-a3b",
"hugging_face_id": "baidu/ERNIE-4.5-21B-A3B-PT",
"source_type": "openrouter_only",
"context_length": 120000,
"max_completion_tokens": 8000,
"is_moderated": false,
"architecture": {
"modality": "text->text",
"input_modalities": [
"text"
],
"output_modalities": [
"text"
],
"tokenizer": "Other",
"instruct_type": null
},
"input_modalities": [
"text"
],
"output_modalities": [
"text"
],
"modality": "text->text",
"tokenizer": "Other",
"instruct_type": null,
"supported_parameters": [
"frequency_penalty",
"max_tokens",
"presence_penalty",
"repetition_penalty",
"seed",
"stop",
"temperature",
"tool_choice",
"tools",
"top_k",
"top_p"
],
"default_parameters": {
"temperature": 0.8,
"top_p": 0.8,
"frequency_penalty": null
},
"per_request_limits": null,
"top_provider": {
"context_length": 120000,
"max_completion_tokens": 8000,
"is_moderated": false
},
"pricing": {
"prompt": "0.00000007",
"completion": "0.00000028"
},
"PPM": {
"prompt": 0.07,
"completion": 0.28
},
"openrouter_raw": {
"id": "baidu/ernie-4.5-21b-a3b",
"canonical_slug": "baidu/ernie-4.5-21b-a3b",
"hugging_face_id": "baidu/ERNIE-4.5-21B-A3B-PT",
"name": "Baidu: ERNIE 4.5 21B A3B",
"created": 1755034167,
"description": "A sophisticated text-based Mixture-of-Experts (MoE) model featuring 21B total parameters with 3B activated per token, delivering exceptional multimodal understanding and generation through heterogeneous MoE structures and modality-isolated routing. Supporting an extensive 131K token context length, the model achieves efficient inference via multi-expert parallel collaboration and quantization, while advanced post-training techniques including SFT, DPO, and UPO ensure optimized performance across diverse applications with specialized routing and balancing losses for superior task handling.",
"context_length": 120000,
"architecture": {
"modality": "text->text",
"input_modalities": [
"text"
],
"output_modalities": [
"text"
],
"tokenizer": "Other",
"instruct_type": null
},
"pricing": {
"prompt": "0.00000007",
"completion": "0.00000028"
},
"top_provider": {
"context_length": 120000,
"max_completion_tokens": 8000,
"is_moderated": false
},
"per_request_limits": null,
"supported_parameters": [
"frequency_penalty",
"max_tokens",
"presence_penalty",
"repetition_penalty",
"seed",
"stop",
"temperature",
"tool_choice",
"tools",
"top_k",
"top_p"
],
"default_parameters": {
"temperature": 0.8,
"top_p": 0.8,
"frequency_penalty": null
},
"expiration_date": null
}
}