Effective from 1 February, the following update will be provisioned in the PolyU GenAI App.
Text Chat Model Refresh
Meta released the Llama -3.3-70B -instruct multilingual language model in December 2024.
The Llama 3.3 instruction-tuned-text-only model is optimized for multilingual dialogue use cases and outperforms many of the available open-source and closed models on common industry benchmarks.
The existing Llama-3.1-70B-instruct model will be replaced by the Llama-3.3-70B-instruct model. Please download and export your chat history from Llama 3.1 if you intend to keep it before 31 January.
The Meta Llama-3.3-70B-instruct model is open-sourced and has been provisioned on-premise to allow staff and students to use it via the PolyU Gen AI app, with no consumption of monthly credit entitlement.
Please refer to the Llama-3.3-70B-Instruct model introduction for further details.
Text to Image Model Refresh
Stability AI released the multimodal diffusion transformer text to image model, Stable Diffusion 3.5 Large, in October 2024.
Stable Diffusion 3.5 Large offers improved performance in image generation quality and supports lengthy text prompts of up to 10K characters. Users can specify a range of sizes for the generated images, with the default being 1024 x 1024 (1:1).
Stable Diffusion 3.5 Large is open-sourced and has been provisioned on-premise to allow staff and students to use it via the PolyU Gen AI app, with no consumption of monthly credit entitlement.
Please refer to the Stable Diffusion 3.5 Large text to image model introduction for further details.
Stability AI released its flagship cloud-based text-to-image model, Stable Image Ultra, in October 2024.
Stable Image Ultra leverage Stable Diffusion 3.5 Large along with advanced workflow to generate the highest quality photorealistic images. Users can input detailed prompts of up to 10K characters and specify a range of sizes, with the default being 1024 x 1024 (1:1). Each image generated by Stable Image Ultra will consume 80 credits. Users are advised to trial and optimize their prompts using the new free on-premise text-to-image models, such as FLUX.1[dev] or Stable Diffusion 3.5 Large, first.
Please refer to the Stable Image Ultra text to image model introduction for further details.
The existing Stable Diffusion XL 1.0 and Stable Diffusion 3 text-to-image models will be replaced by Stable Diffusion 3.5 Large and Stable Image Ultra. Please download and export your chat history from Stable Diffusion XL 1.0 and Stable Diffusion 3 if you intend to keep it before 31 January.
New text-to-image and image-to image model
The Black Forest team released the FLUX.1 [dev] text-to-image, image-to-image model in August 2024.
FLUX.1 [dev] features a 12B parameter rectified flow transformer capable of generating images from text descriptions. The FLUX.1 [dev] model supports image-to-image transformation and modification based on user textual prompt instructions.
FLUX.1[dev] is open-sourced and has been provisioned on-premise to allow staff and students to use it via the PolyU Gen AI app, with no consumption on monthly credit entitlement.
Please refer to Hints & Tips on using Flux.1 [dev] model on text to image and Hints & Tips for using Flux.1 [dev] model to modify previous gen image for further details.
If you need further information or assistance, please contact the IT HelpCentre (Tel: 2766 5900, WhatsApp/ WeChat: 6577 9669) or reach out to us via the IT Online ServiceDesk.
