ERNIE 5.5 Flash Base
Baidu's 280B base model with 14B active and a 1M context, free to test. Its license excludes the EU, UK and South Korea.
In short. ERNIE 5.5 Flash Base is the raw, pre-trained version of a large Baidu language model, meant to be fine-tuned rather than chatted with. It handles very long text and is free to try, but the license does not cover the EU, the UK or South Korea, and it adds revenue-based rules for companies.
Summary
Baidu released ERNIE-5.5-Flash-Base on 5 November 2026. It is the checkpoint at the end of pre-training, before any mid- or post-training, of ERNIE 5.5 Flash, a text-only mixture-of-experts model with 280 billion parameters and 14 billion active per token. It was trained from scratch on 33.5 trillion tokens with a training context of up to 128K, and the released configuration supports 1M tokens for fine-tuning and inference. It has 43 layers, one dense and 42 mixture-of-experts layers with 512 routed experts, 10 chosen per token, plus one shared expert. The attention is a hybrid design: 37 layers of heavily compressed attention, which caches one entry per 128 tokens, and 6 latent-attention layers with sparse attention. Baidu publishes FP8 weights of about 288 GB and BF16 weights of about 573 GB.
As a base model it has no instruction tuning or chat behavior, so it is served through the completions endpoint, and Baidu positions it as a starting point for domain adaptation, instruction tuning and reinforcement-learning research. In Baidu's own table, run through one evaluation pipeline against the base models of DeepSeek-V4-Flash, Hy3-preview and MiMo-V2.5, ERNIE-5.5-Flash-Base leads on most tests, for example MMLU at 93.3 against 89.4, 87.5 and 87.6, and MMLU-Pro at 78.4 against 69.2, 66.9 and 65.7. The same table shows rivals ahead on a few tests, such as DeepSeek on ChineseSimpleQA and Hy3-preview on FullStackBench. All numbers are Baidu's own.
The license is the ERNIE 5.5 Flash Base Model License Agreement v1.0, not Apache or MIT. It is valid only outside the EU, the UK and South Korea. Non-commercial use is free. Commercial use is free up to US$50 million of annual revenue from the related products or services if you show "Powered by ERNIE" and send Baidu a one-time notice, and above that you need Baidu's written authorization. You also may not use the model or its outputs to improve other AI models. The card documents serving with SGLang and vLLM.
Good for
- Researchers and teams that want a large pre-trained mixture of experts to fine-tune or to use for reinforcement-learning research
- Long-context experiments, since the released configuration enables 1M tokens
- Studying a hybrid attention design that mixes heavily compressed and latent attention with multi-router experts
- Companies under US$50 million of related annual revenue that can accept the "Powered by ERNIE" credit and a one-time notice
Watch out for
- It is a base model without instruction tuning, so it does not follow chat instructions out of the box
- The license excludes the EU, the UK and South Korea, and you may not use the model or its outputs to improve other AI models
- Commercial use needs a one-time notice and a "Powered by ERNIE" credit, and above US$50 million of related annual revenue you need Baidu's written authorization
- The FP8 weights are about 288 GB and the BF16 weights about 573 GB, so fine-tuning needs a multi-GPU server
- The benchmark comparison is Baidu's own and covers base models only
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Estimates for the ERNIE-5.5-Flash-Base (FP8) variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.
| Variants | Parameters | BF16 | FP8 / INT8 | 4-bit (Q4) | Fits on |
|---|---|---|---|---|---|
| ERNIE-5.5-Flash-Base (FP8) FP8 | 280B 14B active | 644 GB | 322 GB | 193 GB | Big-memory desktop |
| ERNIE-5.5-Flash-Base-BF16 BF16 | 280B 14B active | 644 GB | 322 GB | 193 GB | Big-memory desktop |
Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.
The license in plain words
Free for non-commercial use, and free for commercial use up to US$50 million of related annual revenue if you show 'Powered by ERNIE' and send Baidu a one-time notice, but only outside the EU, the UK and South Korea. Above US$50 million you need Baidu's written authorization.
- Attribution or model name must be shown
- Free commercial use only below a revenue cap
- Not licensed in some regions
- Acceptable-use policy applies
More about ERNIE 5.5 Flash Base
See also
- Baidu's ERNIE 5.5 Flash Base: 280B parameters, license excludes the EU
- Free until you grow: where open licenses draw the line
Frequently asked questions
- What is ERNIE 5.5 Flash Base?
- ERNIE 5.5 Flash Base is the raw, pre-trained version of a large Baidu language model, meant to be fine-tuned rather than chatted with. It handles very long text and is free to try, but the license does not cover the EU, the UK or South Korea, and it adds revenue-based rules for companies.
- Can I use ERNIE 5.5 Flash Base commercially?
- Yes, with conditions. ERNIE 5.5 Flash Base is released under the ERNIE 5.5 Flash Base, which allows commercial use but adds extra rules: Attribution or model name must be shown; Free commercial use only below a revenue cap; Not licensed in some regions; Acceptable-use policy applies. Read the license text before you ship a product.
- Is ERNIE 5.5 Flash Base open source?
- ERNIE 5.5 Flash Base is open-weight, not open source in the strict sense. You can download the model files, but the ERNIE 5.5 Flash Base adds conditions that the Open Source Initiative definition does not allow.
- Can I run ERNIE 5.5 Flash Base on my own computer?
- Only on a machine with a lot of memory. The smallest variant, ERNIE-5.5-Flash-Base (FP8), needs about 193 GB at 4-bit precision. Desktops with 128 to 512 GB of unified memory can hold it, at modest speed.
- When was ERNIE 5.5 Flash Base released?
- ERNIE 5.5 Flash Base was released by Baidu on November 5, 2026.
- How long a context window does ERNIE 5.5 Flash Base have?
- ERNIE 5.5 Flash Base supports a context window of up to 1M tokens.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- ERNIE 5.5 Flash Base (Baidu ERNIE blog, 5 Nov 2026) ernie.baidu.com
- ERNIE-5.5-Flash-Base model card (Hugging Face) huggingface.co
- ERNIE 5.5 Flash Base Model License Agreement v1.0 huggingface.co
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| DeepSeek V4 DeepSeek | 284B – 1.6T 49B active | 1M | MIT (Permissive) | 176 GB Big-memory desktop | |
| Hy3 Tencent | 295B 21B active | 256K | Apache 2.0 (Permissive) | 204 GB Big-memory desktop | |
| MiMo-V2.6 Xiaomi | 9.4B – 1.02T 42B active | 1M | MIT (Permissive) | 6.49 GB Laptop or phone | |
| Inkling Thinking Machines Lab | 276B – 975B 41B active | 1M | Apache 2.0 (Permissive) | 188 GB Big-memory desktop |