Cheaper than ERNIE 5.0 per token and tuned for agents, which makes it the better Baidu pick for most developers, UK access unconfirmed.
Kenning's take
What it does
ERNIE 5.1 is cut down from ERNIE 5.0, with about a third of the total parameters and half the active ones, which Baidu says cost about 6% of the usual pre-training compute. It was trained further for agent tasks, reasoning and creative writing, and placed 4th on the Arena Search leaderboard on release. It can be used on the ERNIE website and through the Baidu Qianfan API.
- Derived from ERNIE 5.0 at lower cost
- Agent and tool-use training
- 128k token context
- Chat on ernie.baidu.com
- API on Baidu Qianfan (ernie-5.1)
Who it suits. Developers wanting a lower-cost Chinese frontier model, and ERNIE chat users.
Key facts
- Category
- AI models
- Made by
- Baidu
- Based in
- China
- Released
- 9 May 2026
- How to get it
- Chat at ernie.baidu.com; Baidu AI Studio playground; API on Baidu AI Cloud Qianfan
- Works on
- Web, API
- API price
- $0.56 per million input tokens, $2.53 per million output (input up to 32k); $0.84 and $3.10 for 32k to 128k
FAQ
When was ERNIE 5.1 released?
Baidu released ERNIE 5.1 on 9 May 2026.
How much does the ERNIE 5.1 API cost?
$0.56 per million input tokens and $2.53 per million output tokens for inputs up to 32k.
How is ERNIE 5.1 different from ERNIE 5.0?
It has about a third of the parameters of ERNIE 5.0 and focuses on agent tasks, reasoning and writing.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.