Runpod now supports Panchovix’s 16k-token context models, allowing for much deeper context retention in long-form generation. These models require higher VRAM and may trade off some performance, but are ideal for extended sessions like roleplay or complex Q&A.

Runpod announces a partnership with Defined.ai to offer ethically sourced speech and text datasets to AI developers, starting with a pilot program to fine-tune LLMs and accelerate NLP research.

New 8k context models from TheBloke—like WizardLM, Vicuna, and Manticore—allow longer, more immersive text generation in Oobabooga. With more room for character memory and story progression, these models enhance AI storytelling.

Learn how to use Runpod's new Savings Plans to save up to 20% on Secure Cloud pods with monthly or quarterly commitments—ideal for users with high GPU workloads.

Runpod is now an official sponsor of StockDory, a rapidly evolving open-source chess engine that improves faster than Stockfish. StockDory offers deep positional insight, lightning-fast calculations, and full customization—making it ideal for anyone looking to explore AI-driven chess analysis.

While Oobabooga is a popular choice for text-based AI roleplay, KoboldAI offers a powerful alternative with smart context handling, more flexible editing, and better long-term memory retention. This guide compares the two frontends and walks through deploying KoboldAI on Runpod for writers and roleplayers looking for a deeper, more persistent AI interaction experience.

Oobabooga now supports up to 8192 tokens of context, up from the previous 2048-token limit. Learn how to upgrade your install, download compatible models, and optimize your setup to take full advantage of expanded memory capacity in longform text generation.