BaiduLanguage modelPreview
ERNIE 5.5 Flash Preview
Summary
ERNIE 5.5 Flash Preview is a text model with a one-million-token context. Baidu reports 280 billion total parameters, with 14 billion active per token. It builds on ERNIE 5.5 Flash Base and is an intermediate checkpoint from ongoing post-training.
API access is available through Baidu’s Qianfan platform. The weights are still closed. Baidu says it plans to release them later under Apache 2.0.
Similar models
More models of the same kind as ERNIE 5.5 Flash Preview, chosen by model line, developer, size, price and release date.
- ERNIE 5.5 Flash BaseBaidu · same developer
- ERNIE 4.5 300B A47BBaidu · same developer
- ERNIE 4.5 VL 424B A47BBaidu · same developer
- SWE-2 FlashCognition · similar size
- GLM-5.3-FlashZ.ai · similar size
- DeepSeek V4 Flash 0731DeepSeek · similar size
- DeepSeek-V4-FlashDeepSeek · similar size
- Inkling SmallThinking Machines Lab · similar size
All specifications
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
- Access
- Available through the Qianfan API. Qianfan's pricing list names ERNIE-5.5-Flash-Preview as the version, but does not specify a separate HTTP request field.
- Source: Access
- Shutdown
- None announced
- no source
Further details
- Qianfan version name
- Baidu's Qianfan pricing list gives the version name ERNIE-5.5-Flash-Preview. The reviewed page does not publish a separate HTTP request field.
- Source: Qianfan version name
- Checkpoint
- The first post-trained model built on ERNIE 5.5 Flash Base. Baidu describes it as an intermediate checkpoint while post-training continues.
- Source: Checkpoint
- Architecture and size
- 280 billion total parameters, of which 14 billion are active.
- Source: Architecture and size
- Weights and license
- The weights are not available yet. Baidu announces a future release under Apache 2.0.
- Source: Weights and license
Prices
- Standard token rates at peak times
- From 8:00 to 22:00: CNY 0.002 per 1,000 input tokens, CNY 0.0002 per 1,000 cached tokens, and CNY 0.008 per 1,000 output tokens.Source
- Standard token rates off peak
- From 22:00 to 8:00: CNY 0.001 per 1,000 input tokens, CNY 0.0001 per 1,000 cached tokens, and CNY 0.004 per 1,000 output tokens.Source
- Promotional token rates at peak times
- Until November 30, 2026 at 23:59:59: from 8:00 to 22:00, CNY 0.001 per 1,000 input tokens, CNY 0.0001 per 1,000 cached tokens, and CNY 0.004 per 1,000 output tokens. The product page does not state a time zone.Source
- Promotional token rates off peak
- Until November 30, 2026 at 23:59:59: from 22:00 to 8:00, CNY 0.0005 per 1,000 input tokens, CNY 0.00005 per 1,000 cached tokens, and CNY 0.002 per 1,000 output tokens. The product page does not state a time zone.Source
Frequently asked questions about ERNIE 5.5 Flash Preview
How large is the context window of ERNIE 5.5 Flash Preview?
1 M tokens. That is how much text ERNIE 5.5 Flash Preview reads at once, about 1,500 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run ERNIE 5.5 Flash Preview locally?
No. Baidu does not publish the weights.
Stated by Baidu
- Open weights
- no
- License
- n/a
- Parameters
- 280B
- active per token
- 14B
Sources and data date
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved November 12, 2026. |
| Output | Documented modalities. Retrieved November 12, 2026. |
| Additional source | Baidu ERNIE research release index (retrieved November 11, 2026) |
| Additional source | Baidu Qianfan model access (retrieved November 12, 2026) |
| Additional source | Baidu Qianfan model overview (retrieved November 12, 2026) |
| Additional source | Baidu Qianfan token pricing (retrieved November 12, 2026) |