ERNIE X1 Turbo

Baidu’s ERNIE X1 Turbo (launched Apr 25, 2025 at Baidu Create) targets complex reasoning with a more advanced chain-of-thought, improved deep-thinking behavior, and upgraded multimodal + tool-invocation skills. Baidu’s launch materials say it excels at Q&A, literary creation, and logical reasoning, and claims superiority over DeepSeek R1/V3. It’s also positioned as faster and cheaper: half the price of ERNIE X1 (about ¥1 per million input tokens and ¥4 per million output).
For access, developers can use ERNIE Bot and Baidu’s Qianfan API. Qianfan lists multiple endpoints, including ernie-x1-turbo-32k (32K context; up to ~28K output; high throughput around 900 RPM/300k TPM) and ernie-x1-turbo-latest (64K context). These entries also note that X-series “deep thinking” models don’t support adjustable thinking budgets and that max_tokens covers both reasoning and final content.
There’s also a public demo workspace for ERNIE X1 Turbo showcasing its interactive behavior. If you need a low-cost, high-throughput reasoning model with solid multimodal/tool abilities and generous context, X1 Turbo is Baidu’s value-oriented pick.