Baidu: ERNIE 4.5 21B A3B

Open ERNIE text MoE with 21B total and 3B active parameters

Model lab: Baidu
Model family: ernie

Model resources

Hugging Face

Modalitiestext → text

Context window131,072

Maximum output8,000

Input · Context ≤ 120000 tokens · Maximum output 8000 tokensUSD 0.056 / 1M tokens

Output · Context ≤ 120000 tokens · Maximum output 8000 tokensUSD 0.224 / 1M tokens

Try now View docs

Quickstart

Use your Moduck API Key, set the API base URL to https://api.moduck.ai/v1, and set the model parameter to baidu/ernie-4.5-21b-a3b.

import os
import requests

base_url = "https://api.moduck.ai/v1"
api_key = os.environ["MODUCK_API_KEY"]
model_id = "baidu/ernie-4.5-21b-a3b"

response = requests.post(
    f"{base_url}/chat/completions",
    headers={
        "Authorization": f"Bearer {api_key}",
        "Content-Type": "application/json",
    },
    json={
        "model": model_id,
        "messages": [{"role": "user", "content": "Hello!"}],
        "stream": False,
    },
    timeout=60,
)
response.raise_for_status()
print(response.json())

Use the familiar OpenAI-style chat request format to make model integration easier.

ERNIE 4.5 21B A3B API Pricing

Type	Price information
Input · Context ≤ 120000 tokens · Maximum output 8000 tokens	USD 0.056 / 1M tokens
Output · Context ≤ 120000 tokens · Maximum output 8000 tokens	USD 0.224 / 1M tokens

Performance

Performance trends

Throughput88.5 tok/sCurrent

Time to first token1.4 sCurrent

End-to-end latency8.0 sCurrent

Throughput

Current88.5 tok/s

Window average89.1 tok/s

Time to first token

Current1.4 s

Window average1.5 s

End-to-end latency

Current8.0 s

Window average7.5 s

Tail latency (P99)

Current15.2 s

Window average16.9 s

Prompt cache hit rate

Current18.1%

Window average24.6%

Error rate

Current0.0%

Window average0.2%

Availability

Uptime100%At least one provider responded

Availability100%Requests actually served

Error rate0%Share of failed requests

Served requests, hour by hour

24 hours agoNow

ERNIE 4.5 21B A3B Compare models

Model	Input price	Output price	Context window	Capabilities
ERNIE 4.5 21B A3BThis page	USD 0.056 / 1M tokens · Context ≤ 120000 tokens · Maximum output 8000 tokens	USD 0.224 / 1M tokens · Context ≤ 120000 tokens · Maximum output 8000 tokens	131,072	Text output · Tool calling
ERNIE 4.5 VL 424B A47B	USD 0.336 / 1M tokens	USD 1 / 1M tokens	123,000	Text output · Reasoning
CoBuddy	USD 0.224 / 1M tokens	USD 0.904 / 1M tokens	131,072	Text output · Reasoning · Tool calling
Beam 1.0 501B A23B	USD 0.2 / 1M tokens	USD 0.8 / 1M tokens	1,048,576	Text output · Reasoning · Tool calling
Ling 3.1 Flash	USD 0.06 / 1M tokens	USD 0.176 / 1M tokens	262,144	Text output · Reasoning · Tool calling

Model specifications

Input: text
Output: text
Context window: 131,072
Maximum output: 8,000
Release date: 2025-06-30
Attachments: No
Reasoning: No
Tool calling: Yes
Temperature setting: Yes
Open weights: Yes

Frequently asked questions

What is ERNIE 4.5 21B A3B for?

Open ERNIE text MoE with 21B total and 3B active parameters

What inputs and outputs does ERNIE 4.5 21B A3B support?

Recorded inputs: text; outputs: text.

How is ERNIE 4.5 21B A3B priced?

The pricing table shows the current recorded Moduck price. For example: Input · Context ≤ 120000 tokens · Maximum output 8000 tokens USD 0.056 / 1M tokens; Output · Context ≤ 120000 tokens · Maximum output 8000 tokens USD 0.224 / 1M tokens. See the pricing table for all units, variants, and conditions.

What specifications are recorded for ERNIE 4.5 21B A3B?

Current model records list Context window 131,072 tokens; Maximum output 8,000 tokens.

What model parameter does the Quickstart use?

The example uses baidu/ernie-4.5-21b-a3b. This value comes from the page identifier and is not confirmed as a callable Moduck public ID.

Can I call this model through the Moduck API now?

This page has no verified Moduck API call for this model. A Quickstart example or reserved ID does not establish access; check the models available to your account.