The latest AI news and updates of Ollama — Open-source tool for running, managing, and serving large language models locally. Covering Ollama's latest product updates from the past 90 days.
OllamaJul 27Ollama added Moonshot AI’s Kimi K3, a 2.8-trillion-parameter multimodal model, to its cloud platform. The model features a 1-million-token context window and native visual understanding. It is available to Pro and Max subscribers via extra usage credits and integrates with agentic coding tools including Claude Code, OpenCode, Hermes Agent, and OpenClaw.
OllamaJul 17Ollama 0.32.1 updates Gemma 4 with improved tool calling reliability for coding agents. The release addresses previous consistency issues, ensuring more dependable performance in autonomous workflows. The 26B model runs via the Pi agent, with an optional MLX engine available for maximum performance on Apple Silicon.
OllamaJul 15Ollama now integrates with OpenCode Desktop, enabling the terminal-based coding agent to run with local or cloud models. The integration launches via the ollama launch opencode command and supports file editing, command execution, web fetching, and vision capabilities. Models require a context window of 64k tokens or higher for repository-wide tasks.
OllamaJul 8Ollama increased cloud capacity for Z.ai’s GLM-5.2 model across US and European regions. The infrastructure update delivers 80 to 120 output tokens per second, significantly outpacing the 30 to 40 tokens per second reported on other providers. The model remains accessible for agentic coding workflows via Claude Code, VS Code, Codex, and Hermes using the ollama launch command.
OllamaJul 1Ollama 0.31 enables multi-token prediction by default for Gemma 4 on Apple Silicon, increasing generation speed by nearly 90%. The engine uses a small draft model to propose tokens, which the main model verifies in a single pass. This optimization, measured at 95.0 tokens per second on an M5 Max, accelerates agentic coding tasks without requiring manual configuration.
OllamaJun 28Ollama adds the Ornith-1.0 family of open-source agentic coding models to its library. The models, ranging from 9B to 397B parameters, feature a 256K context window and use reinforcement learning for self-improving code generation. The integration supports direct execution via ollama run and agentic workflows through ollama launch for Claude Code and Pi.
OllamaJun 16Ollama is now hosting Z.ai's GLM-5.2 model on its US-based cloud, powered by NVIDIA Blackwell GPUs. The model features a 1-million-token context window and two reasoning effort levels for long-horizon coding tasks. It is available for immediate use within Claude Code, Codex App, and Hermes Agent via the ollama launch command, with a zero-data-retention privacy policy.
OllamaJun 15Ollama now supports Cline CLI, enabling the autonomous coding agent to launch directly from the terminal. The integration auto-configures local model providers and supports the Kanban feature, which runs multiple coding agents in parallel to handle separate project tasks simultaneously. Sessions launch with a chosen local or cloud model via the ollama launch cline command.
OllamaJun 15Ollama added Moonshot AI's Kimi K2.7 Code model to its US-hosted cloud on NVIDIA B300 GPUs. The model supports text and image input with a 256K context window. It is available for immediate use within Claude Code, Codex App, and OpenCode via the ollama launch command, with no user data retention or training.
OllamaJun 10Ollama now supports Nous Research's Hermes Desktop, enabling users to run the multi-agent system locally or in the cloud. This integration brings Hermes Desktop's self-improving AI agents and messaging capabilities to Ollama's local model deployment platform. It allows users to manage complex agentic workflows with greater control over their compute environment.
OllamaJun 7Ollama has made NVIDIA's Nemotron 3 Ultra model available on its cloud. This 550 billion parameter Mixture of Experts (MoE) model is designed for long-running AI agents, delivering 5x faster inference and up to 30% lower costs for complex agentic tasks.
OllamaJun 7Ollama has made Google DeepMind's Gemma 4 12B model available for local execution, including support for chat and agentic applications. This expands access to a powerful, open-weight multimodal model optimized for on-device reasoning and coding, enabling private and offline AI workflows on consumer hardware.
OllamaJun 7Ollama has made the MiniMax M3 model available on its Cloud, providing US-based access with zero data retention. This integration offers a frontier-level, open-weight model for agentic coding and multimodal tasks, featuring a 1-million-token context window. It expands access to advanced AI capabilities for complex, autonomous workflows.
Frequently asked questions