LIVE

Publisher profile

baidu AI models

Explore active AI models attributed to baidu across indexed source catalogs. Open any profile for capabilities, context, pricing and verified source links.

Models
6
Categories
2
Sources
2
reasoningvision

Available catalog

Published models

6 active

OpenRouter↗

Baidu: ERNIE 4.5 VL 424B A47B

ERNIE-4.5-VL-424B-A47B is a multimodal Mixture-of-Experts (MoE) model from Baidu’s ERNIE 4.5 series, featuring 424B total parameters with 47B active per token. It is trained jointly on text and image data...

reasoningvisiontextimage
Size
424B parameters
Context
123,000
OpenRouter↗

Baidu: ERNIE-5.5-Flash-Preview

Baidu: ERNIE-5.5-Flash-Preview is a text-only preview foundation model intended for coding, agentic workflows, knowledge work, reasoning, and deep research. The supplied OpenRouter material describes it as the latest model in Baidu’s ERNIE family, with advertised improvements in coding, agentic capabilities, knowledge work, and deep research. Tool-calling is listed as a capability, making it relevant for applications that need external tools or multi-step agent behavior. Its advertised context window is 1,024,000 tokens, which may suit long documents, extended research tasks, and large code-related inputs; the supplied material does not specify a separate runtime default. OpenRouter lists pricing of $0.28 per million input tokens and $1.13 per million output tokens. The model record does not provide a verified parameter count, inference-speed benchmark, license, model task classification, or quantization details. Accordingly, the 280-billion figure mentioned in the existing summary should not be treated as a confirmed parameter count for this record. Because this is labeled Preview, production behavior and provider availability may change, and the supplied material does not include benchmark results or additional safety documentation.

reasoningtext
Size
Size not listed
Context
1,024,000
OpenRouter↗

Baidu: ERNIE-5.5-Flash-Preview (free)

Baidu: ERNIE-5.5-Flash-Preview (free) is a text-only preview foundation model available through OpenRouter at no listed input or output cost. The supplied material describes it as the latest model in Baidu’s ERNIE family, with advertised improvements in coding, agentic capabilities, knowledge work, and deep research. Its practical fit is therefore reasoning-heavy text work, research assistance, coding, and agent workflows that need tool-calling support. The advertised context window is 1,024,000 tokens, which is unusually large for processing long prompts or documents; the supplied material does not provide a separate runtime default or configuration. The model is marked as supporting tool-calling, but no details are provided about supported tools, structured-output behavior, latency, throughput, or reliability. It is identified as a preview release, so behavior and availability may change. Parameter count, quantization, model task, license, and an inference speed benchmark are not available in the supplied material. The record does not establish whether the free offering has usage limits or differs technically from the underlying base model beyond its OpenRouter route.

reasoningtext
Size
Size not listed
Context
1,024,000
Hugging Face↗

ERNIE-5.5-Flash-Base

ERNIE-5.5-Flash-Base is a Hugging Face model repository for text generation with 286.2B parameters under the other license.

text
Size
286.2B parameters
Context
Not listed
Hugging Face↗

Unlimited-OCR

Unlimited-OCR is a vision-oriented Hugging Face model for converting image content into text output. The supplied model page classifies it as an “Image-Text-to-Text” model and describes it as having 3B parameters, while the model record specifies an exact parameter count of 3,336,106,240. Its listed modalities are image and text, making it potentially relevant for OCR workflows, document text extraction, and applications that need to interpret visual input and return textual results. The repository is available through Hugging Face and is identified in the model record as using the MIT license, which is a permissive licensing caveat to review alongside any repository-specific terms or included assets. No specific task capabilities, supported languages, accuracy results, latency measurements, inference hardware requirements, or deployment examples are available in the supplied material. The context window is not available, and neither input nor output pricing is provided. A benchmarked inference speed is also unavailable, so performance and infrastructure fit should be validated experimentally before production use. The source page provides files and model exploration through Hugging Face, but the supplied excerpt does not establish an API endpoint or hosted inference offering.

visiontextimage
Size
3.3B parameters
Context
Not listed
Hugging Face↗

Unlimited-OCR-2

Unlimited-OCR-2 is an image-text-to-text model repository on Hugging Face intended for workflows that accept images and produce text. Its listed modalities are image and text, and the supplied record assigns it to the vision use case. This makes it potentially relevant for visual document or OCR-oriented pipelines, although the supplied material does not provide OCR accuracy results, supported languages, example outputs, or task-specific benchmarks. The record reports 3,343,978,497 parameters, while the Hugging Face page describes it more generally as a model with 3B parameters. No quantization is listed, so this record should be treated as describing the base model rather than a quantized upload. The license for this record is MIT. A context window, inference speed, pricing, hardware requirements, implementation examples, and detailed capability tags are not available in the supplied material. In particular, there is no verified tokens-per-second benchmark or advertised context limit to use for capacity planning. Developers should therefore validate image resolution support, text-generation behavior, deployment requirements, and OCR quality against their own inputs before adopting it.

visiontextimage
Size
3.3B parameters
Context
Not listed

Continue exploring

Related guides and tools