Mesolitica’s Post

Hi everyone, Mesolitica launched MaLLaM 🌙 (Malaysia Large Language Model) API! Multi-lingual Malaysian Chat Language Model, 32k context length, Malaysian centric and private, https://lnkd.in/gy9cp9mr 1. Natively understand Standard Malay, local Malay, Jawi, Standard English, Manglish, Mandarin and Indonesian as input. Standard English and Manglish will always reply in English, others will always reply in Malay. 2. Natively understand Retrieval-Augmented Generation, aka RAG. You just give prefix document and you can start your typical RAG QA sessions. 3. Natively understand Function Call, you can convert human natural text to function parameters and call the appropriate function, 100% done by LLM end-to-end. 4. Better accuracy than Mesolitica Open Models and ChatGPT3.5 using Tatabahasa and PT3 benchmarks. 5. You can play around with MaLLaM 🌙 inside Nous Playground. It support function call, multi-turn and export code, also you can check the cookbooks to help your development. AI Safety is coming soon which is free of charge part of the API! 6. The API is compatible with Python OpenAI library, or if you prefer cURL. 7. Our pricing is competitive with available LLM APIs out there, existing users or first registration will get free RM2.5 credit! 8. You can topup your MaLLaM 🌙 tokens at billing page, for now we only support Stripe. See picture 5. 9. We also open for self-host enterprise MaLLaM 🌙 in your private cloud or on-premise, 100% internet is not required, we collaborate with Nvidia APAC to include hardware support, read more about our self-host enterprise package at https://lnkd.in/giPmVaEj 10. If you are a researcher and interested to study MaLLaM🌙 extensively such as alignment, accuracy and etc, we can provide certain free credits, which will benefit for both parties, feel free to PM me. 11. We already started working on Code Interpreter, https://lnkd.in/gShK496m, LLM agent to execute code on your behalf, it can be your data analyst or python programmer companion, will be release in the next version. 12. Open source still at our heart, this is necessary to upskill the others, you can checkout our massive open malaysian models and dataset at https://lnkd.in/eU9KYafq We are looking forward for the feedback and the development of Malaysian Large Language Model for better tech!