ERNIE 4.5 21B A3B Thinking
ERNIE 4.5 21B A3B Thinking is a chat model from Baidu, released October 9, 2025. ERNIE-4.5-21B-A3B-Thinking is Baidu's upgraded lightweight MoE model, refined to boost reasoning depth and quality for top-tier performance in logical puzzles, math, science, coding, text generation, and expert-level academic benchmarks.
by Baidu · 131K token context window
Best for
- low-latency chat and routing
- request routing and triage
- text classification
Ways to use ERNIE 4.5 21B A3B Thinking in osFoundry
Connect with your own key (BYOK)
Open the key dialog and paste your Baidu API key. osFoundry discovers ERNIE 4.5 21B A3B Thinking automatically — assign it to a Maestro role (router, direct, orchestrator, or fallback) in the Pipeline tab and it is live in every chat. Your key, your provider account — no token markup.
Use it in a Room App
Room Apps declare AI features in their manifest, then call them with invokeAI:
import { invokeAI } from '@osfoundry/app-sdk'
// 'summarize' is an AI feature declared in your app manifest.
const result = await invokeAI('summarize', userText)
Call it from your own apps
Once a model is wired into your workspace you can host it as an API and reach it from your own services, scripts, or CI — outside osFoundry.
Run ERNIE 4.5 21B A3B Thinking yourself
ERNIE 4.5 21B A3B Thinking is also available as open weights — self-host it for full data control and no per-token cost. See that page for GPU requirements and a cost comparison against API pricing.
ERNIE 4.5 21B A3B Thinking vs similar models
Licence
Hosted — usage subject to provider terms — Hosted-only model — usage governed by the provider's API terms. Bring your own provider key.
No weights distributed; usage subject to provider terms.
Frequently asked about ERNIE 4.5 21B A3B Thinking
How much does ERNIE 4.5 21B A3B Thinking cost?
ERNIE 4.5 21B A3B Thinking is metered at $ 0.070 /1M for input, and $ 0.280 /1M for output. Bring your own Baidu API key — osFoundry passes through provider pricing without markup.
Can I use ERNIE 4.5 21B A3B Thinking commercially?
Commercial use is allowed with conditions. Hosted-only model — usage governed by the provider's API terms. Bring your own provider key. No weights distributed; usage subject to provider terms.
What is the context window of ERNIE 4.5 21B A3B Thinking?
ERNIE 4.5 21B A3B Thinking supports a 131K token context window.
Can I run ERNIE 4.5 21B A3B Thinking locally?
No — ERNIE 4.5 21B A3B Thinking is hosted only and accessed via the Baidu API. An open-weights equivalent is available to self-host — see the cross-link above.
What is ERNIE 4.5 21B A3B Thinking best at?
ERNIE 4.5 21B A3B Thinking is well-suited to low-latency chat and routing, request routing and triage, text classification.
How do I use ERNIE 4.5 21B A3B Thinking in osFoundry?
Paste your Baidu API key in the key dialog (or deploy the open weights for self-hostable models), assign ERNIE 4.5 21B A3B Thinking to a Maestro role in the Pipeline tab, then use it in chat, Room Apps via invokeAI, or your own apps.
Published by Baidu on October 9, 2025. Source: https://openrouter.ai/baidu/ernie-4.5-21b-a3b-thinking