> For the complete documentation index, see [llms.txt](https://changelog.graphlit.dev/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://changelog.graphlit.dev/september-2024/september-26-support-for-google-ai-and-cerebras-models-and-latest-groq-models.md).

# September 26: Support for Google AI and Cerebras models, and latest Groq models

### New Features

* :bulb: Graphlit now supports the [Cerebras](https://cerebras.ai/) model service which offers the `LLAMA_3_1_70B` and `LLAMA_3_1_8B` models.
* :bulb: Graphlit now supports the [Google AI](https://ai.google.dev/) model service which offers the `GEMINI_1_5_PRO` and `GEMINI_1_5_FLASH` models.
* We have added support for the latest Groq Llama 3.2 preview models, including `LLAMA_3_2_1B_PREVIEW`, `LLAMA_3_2_3B_PREVIEW`, `LLAMA_3_2_11B_TEXT_PREVIEW`, and `LLAMA_3_2_90B_TEXT_PREVIEW`.  We have also added support for the Llama 3.2 multimodal model `LLAMA_3_2_11B_VISION_PREVIEW.`
* We have added a new `specification` parameter to the `promptConversation` mutation. Now you can specify your initial specification for a new conversation, or update an existing conversation, without requiring additional API calls.
* :zap: We have changed the retrieval behavior of the `promptConversation` mutation. Now, if no relevant content was found via vector-based semantic search (given the user prompt), we will fallback to any relevant content from the message in the conversation. If there was no content from the conversation to fallback to, we will fallback to the last ingested content in the project. This solves an issue where a first prompt like 'Summarize this' would find no relevant content.  Now it will fallback to retrieve the last ingested content.
* :zap: We have renamed the Groq model enum from `LLAVA_1_5_7B` to `LLAVA_1_5_7B_PREVIEW.`

### Bugs Fixed

* GPLA-3083: Not sending custom instructions/guidance with extraction prompt
* GPLA-3146: Filtering Persons by email not working
* GPLA-3171: Not failing on deprecated OpenAI model
* GPLA-3158: Summarization not using revision strategy