[GH-ISSUE #6193] Add New SOTA Models: Palmyra-Med and Palmyra-Fin #65903

Closed
opened 2026-05-03 23:07:23 -05:00 by GiteaMirror · 5 comments
Owner

Originally created by @gileneusz on GitHub (Aug 6, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6193

Request generated by ChatGPT cause I don't have time to write such posts and I'm lazy too. Thanks for understanding my lazy nature.

I recently came across two impressive state-of-the-art (SOTA) domain-specific models, Palmyra-Med and Palmyra-Fin, described on https://writer.com/blog/palmyra-med-fin-models/. I believe these models would be a valuable addition to the Ollama platform. Here’s a brief overview:

Palmyra-Med: https://huggingface.co/Writer/Palmyra-Med-70B-32K

Designed for medical applications.
Achieves 85.9% average across all medical benchmarks, surpassing Med-PaLM-2.
Excels in clinical knowledge (90.9% in MMLU Clinical Knowledge) and anatomy (83.7% in MMLU Anatomy).
Supports diagnostic accuracy, treatment planning, genetic counseling, and biomedical research.
Cost-effective at $10 per 1M output tokens, compared to $60 for GPT-4.

Palmyra-Fin: https://huggingface.co/Writer/Palmyra-Fin-70B-32K

Tailored for financial applications.
Supports financial trend analysis, investment analysis, risk evaluation, and asset allocation strategy.
Utilizes well-curated financial training data and fine-tuning instruction data to ensure high accuracy.
Both models integrate seamlessly with the Writer full-stack generative AI platform, offering tools like integrated graph-based RAG technology, AI guardrails, and a suite of developer tools. Available via API, No-code tools, and the Writer Framework, they come with an open-model license for easy deployment locally or in private clouds.

Adding these models to Ollama would significantly enhance the platform's capabilities in the medical and financial sectors, providing users with highly specialized and accurate AI tools.

Thank you for considering this request!

Originally created by @gileneusz on GitHub (Aug 6, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6193 Request generated by ChatGPT cause I don't have time to write such posts and I'm lazy too. Thanks for understanding my lazy nature. I recently came across two impressive state-of-the-art (SOTA) domain-specific models, Palmyra-Med and Palmyra-Fin, described on https://writer.com/blog/palmyra-med-fin-models/. I believe these models would be a valuable addition to the Ollama platform. Here’s a brief overview: Palmyra-Med: https://huggingface.co/Writer/Palmyra-Med-70B-32K Designed for medical applications. Achieves 85.9% average across all medical benchmarks, surpassing Med-PaLM-2. Excels in clinical knowledge (90.9% in MMLU Clinical Knowledge) and anatomy (83.7% in MMLU Anatomy). Supports diagnostic accuracy, treatment planning, genetic counseling, and biomedical research. Cost-effective at $10 per 1M output tokens, compared to $60 for GPT-4. Palmyra-Fin: https://huggingface.co/Writer/Palmyra-Fin-70B-32K Tailored for financial applications. Supports financial trend analysis, investment analysis, risk evaluation, and asset allocation strategy. Utilizes well-curated financial training data and fine-tuning instruction data to ensure high accuracy. Both models integrate seamlessly with the Writer full-stack generative AI platform, offering tools like integrated graph-based RAG technology, AI guardrails, and a suite of developer tools. Available via API, No-code tools, and the Writer Framework, they come with an open-model license for easy deployment locally or in private clouds. Adding these models to Ollama would significantly enhance the platform's capabilities in the medical and financial sectors, providing users with highly specialized and accurate AI tools. Thank you for considering this request!
GiteaMirror added the model label 2026-05-03 23:07:23 -05:00
Author
Owner

@gileneusz commented on GitHub (Aug 7, 2024):

there is support from the authors to add it to ollama

Screenshot 2024-08-08 at 00 39 35
<!-- gh-comment-id:2274469601 --> @gileneusz commented on GitHub (Aug 7, 2024): there is support from the authors to add it to ollama <img width="803" alt="Screenshot 2024-08-08 at 00 39 35" src="https://github.com/user-attachments/assets/2bc1b6a6-2f50-4614-8394-46a5be5b9a36">
Author
Owner

@gileneusz commented on GitHub (Aug 13, 2024):

@igorschlum any chances for those models to be added? 🥹 🥺

<!-- gh-comment-id:2286736730 --> @gileneusz commented on GitHub (Aug 13, 2024): @igorschlum any chances for those models to be added? 🥹 🥺
Author
Owner

@igorschlum commented on GitHub (Aug 13, 2024):

hi @gileneusz I'm a user of Ollama as you. I'm not in charge of uploading models on the librairy. I'm seen that you need to accept a licence to use Palmyra-Med and that the licence is for a non commercial use only. I think that you made the good choice to ask them to upload it to Ollama website.

<!-- gh-comment-id:2287140783 --> @igorschlum commented on GitHub (Aug 13, 2024): hi @gileneusz I'm a user of Ollama as you. I'm not in charge of uploading models on the librairy. I'm seen that you need to accept a licence to use Palmyra-Med and that the licence is for a non commercial use only. I think that you made the good choice to ask them to upload it to Ollama website.
Author
Owner

@tm17-abcgen commented on GitHub (Aug 16, 2024):

Added the models and quoted your request @gileneusz to Ollamas Model Repository. Q4_K_M only though. Can be found here:
https://ollama.com/thewindmom/palmyra-med-70b-32k

<!-- gh-comment-id:2294103970 --> @tm17-abcgen commented on GitHub (Aug 16, 2024): Added the models and quoted your request @gileneusz to Ollamas Model Repository. Q4_K_M only though. Can be found here: https://ollama.com/thewindmom/palmyra-med-70b-32k
Author
Owner

@gileneusz commented on GitHub (Aug 17, 2024):

@tm17-abcgen wow thank you so much 🤗 can you also add more quants later? like Q8_0, Q6_0? or fp16 for better accuracy in benchmarks?

<!-- gh-comment-id:2294793722 --> @gileneusz commented on GitHub (Aug 17, 2024): @tm17-abcgen wow thank you so much 🤗 can you also add more quants later? like Q8_0, Q6_0? or fp16 for better accuracy in benchmarks?
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#65903