# Baidu says ERNIE 5.5 Flash Preview has launched, with weights due November 25

> Baidu says its ERNIE 5.5 Flash Preview has launched, and model weights are expected by November 25.

- Status: reported (One outlet, not independently verified yet.)
- Topic: Models
- Published: 2026-11-13T09:05:56+00:00
- Updated: 2026-11-14T09:09:25.369816+00:00
- Web page: https://ai-news.ikangai.com/story/baidu-says-ernie-5-5-flash-preview-has-launched-with-weights-due-november-25/
- Seen in: Latent Space
- Change: capability (New capability)
- Applies to: ERNIE 5.5 Flash Preview

## What changed

Baidu has launched ERNIE 5.5 Flash Preview, and its model weights are expected by November 25.

Baidu has announced that ERNIE 5.5 Flash Preview has launched and that its model weights are expected by November 25, according to a screenshot of the announcement shared on Reddit. The announcement describes a model with 280B total parameters and 14B active parameters, a 1M-token context window, and generation above 200 tokens per second. These are claims made in the announcement and have not been independently verified. The model is positioned for coding, 3D modeling, knowledge work, and content creation. The post does not say what hardware the 200-plus tokens per second figure was measured on.

Commenters on the Reddit thread were cautiously positive that Baidu gave a firm date rather than a vague promise to release soon. Some said Baidu has underperformed despite its early machine-learning research. One commenter joked that they would never reach 200 tokens per second running it locally.

The same commenter offers a rough sizing estimate: a Q4-class or at least IQ4_XS quantization may suit a system with about 32GB of VRAM and 128GB of system RAM. They also link a draft llama.cpp pull request, ggml-org/llama.cpp#31925, which suggests community work is already underway toward running the model in that stack.

## Why it matters

A large long-context model from a major Chinese lab could give teams another open-weight option for coding and document work, if the weights and claims hold up.

## What you can do

Wait for the weights and independent benchmarks before testing it in production.

## How the story developed

- 2026-11-14: Adds the stated 280B total and 14B active parameters, 1M-token context, free OpenRouter access, and the Apache 2.0 licence cited by Arena.

## Further reporting

- [@omarsar0](https://link.mail.beehiiv.com/ss/c/u001.r2tkAX6RdgQsMKMmEZrF6LfzwOLJP3GDa3E_YvIcF9S9Uiiip-QG58yvpZJkhqa-prx7EfjER42S5XIc9r72XAOySJQ_cM2d2ebQUMEcETazyetTmOVrBGedbDiBwZAQFazmaNKSTJJSJOUL8ezJMw/4vq/NuIWVQqdQ0W3SKogoWyaGg/h31/h001.rI9L7VlMLW7TrIvziBn6NHBFMhVt-6I7D-c14-kYTqg)

---

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it.

Cite as: IKANGAI AI News, "Baidu says ERNIE 5.5 Flash Preview has launched, with weights due November 25", 2026-11-13, https://ai-news.ikangai.com/story/baidu-says-ernie-5-5-flash-preview-has-launched-with-weights-due-november-25/
