News · Baidu · 2 min read
Event date: · Published:
Baidu previews ERNIE 5.5 Flash with 1M context, plans Apache 2.0 release
Baidu released ERNIE 5.5 Flash Preview on November 11, 2026, a 280 billion parameter MoE model with a 1 million token context window and a promised Apache 2.0 open weights release.
Baidu released ERNIE 5.5 Flash Preview on November 11, 2026, calling it the first post-trained version of its ERNIE 5.5 Flash Base checkpoint. The announcement appeared on ERNIE 5.5 Flash Preview, Baidu's official release page for the model.
What changed in this release
- The model has 280 billion total parameters with 14 billion active parameters in a sparse mixture-of-experts design.
- It supports a 1 million token context window.
- Baidu says the model writes code, uses tools, and iterates to build games, create 3D models, and produce videos.
- The model is live now via Baidu's Qianfan platform and listed on OpenRouter.
- Current weights are released under the ERNIE 5.5 Flash Base community license, with Baidu saying it will open-source the full weights under Apache 2.0 soon.
What it means for people who build with AI
For vibecoders and tool builders, a 1 million token context window paired with code generation and tool use gives more room to work with large codebases, long conversations, or multi-step agent tasks in a single pass. Baidu's promise of full Apache 2.0 open weights would turn ERNIE 5.5 Flash into another large open option to evaluate once those weights ship, alongside the proprietary alternatives most builders already use.
Availability and what's next
Baidu describes ERNIE 5.5 Flash Preview as an intermediate checkpoint, noting that training is not finished and that some capabilities may still fall short. The model is accessible now through Qianfan and OpenRouter under the community license, while the fully open Apache 2.0 weights are still to come.
Sources
- ERNIE 5.5 Flash Preview (Baidu, November 11, 2026, official)
Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.