Adding a flagship open-weight model like Kimi K3 to Venice’s Agentic Chat would completely skyrocket the platform's capabilities for deep research and coding. Right now, there is a massive capability gap between expensive pay-per-token models (like 5.6-Terra or Opus 5) and private, tokenless LLMs. Because Kimi K3 is an open-weight model, if Venice had the backend infrastructure to host it, they could offer a model that literally scored a 93.5 on the GPQA Diamond benchmark as an agent.
(and it’d be really cool if we got more models that didn’t charge per token in Agentic Mode in general)