Ollama Adds New Model Support and Features

What happened

Ollama has expanded its model support to include Xiaomi MiMo's mimo-v2.6, a 9B model designed for code and agent workflows. The AutoTrust decision models are also now available, with some models capable of analyzing text and providing choices, yes or no probabilities, and ratings for questions.

The release also introduces support for Kimi Code CLI, which is the successor to kimi-cli, and enables migration for existing kimi-cli setups. Additionally, Ollama automatically fills in Pi's model list, allowing models to be selected within Pi, and Claude Desktop's model picker now displays Ollama's actual model names.

Why it matters

These updates enhance Ollama's versatility by integrating models with specialized capabilities, such as code generation and decision analysis. The inclusion of AutoTrust models with image analysis features expands the platform's utility for developers and researchers working on complex tasks.

Key details

The AutoTrust family includes models with 9B, 35B, and 27B-vl sizes, with the 35B and 27B-vl sizes also capable of reading images. The autotrust-jev model is now supported on Apple silicon, making it compatible with Apple's hardware ecosystem.

Ollama also supports recurrent and vision GGUF models handling multiple requests simultaneously, improving efficiency for users. Qwen3 and Qwen3.5 models are now using long-context (YaRN) scaling, which may enhance their performance on extended input tasks.