Ollama
About Ollama
Ollama develops software for running and connecting to language models. Its local runtime downloads models for execution on a user's computer, while Ollama Cloud provides hosted model access. The project also maintains a model library, APIs and client libraries for applications, editors and coding agents. A model listed in Ollama can originate from another developer; the runtime, distribution entry and underlying model are separate parts of an observation.
Focus
- Local model execution and management
- Hosted cloud model access
- Model distribution and application integrations
Projects and interfaces
Ollama runtime
runtime
Runs downloaded models locally and exposes interfaces used by applications and coding tools.
Model library
model catalog
Lists model packages for tasks including text, coding, vision, embeddings and reasoning, with release-specific model information.
Ollama Cloud
hosted service
Provides hosted model execution through apps, terminal workflows and authenticated APIs.
API and official client libraries
developer interface
Documents native requests and compatible client interfaces, with official Python and JavaScript libraries.
Tool calling
documented capabilities
Documents how a model requests a tool call and how the application supplies the tool result in a follow-up exchange.
Structured outputs
documented capability
Describes JSON-schema response formats; the documented cloud restriction is stated separately.
Embeddings
documented capability
Generates vector representations of text for retrieval and related application tasks.
Scope and limitations
- A library listing identifies a distribution route; model authorship and licensing belong to the individual release.
- The structured-output documentation checked on this date states that this feature is not supported by Ollama Cloud.
- Local and cloud execution have different data paths and availability conditions.
Official references
- Ollama documentation. Checked .
- Ollama Cloud. Checked .
- Ollama API introduction. Checked .
- Ollama model library. Checked .
- Tool calling. Checked .
- Structured outputs. Checked .
- Embeddings. Checked .
Harnesses & editors
- Ollama
Observations
- 14B-32B local models — all 47 · V1
- Agent harnesses and tooling · V2
- Aider — 169 runs across 83 models · V1
- Architecture census — 271 models scanned · V1
- Cloud Models — three-judge autoresearch observation set · V1
- NVIDIA GTX 1070 — 344 models profiled · V1
- Ollama deployment composite observation · V1
- Pre-retirement tool-use observation: three models through Ollama cloud · V1
- Sub-3B local models — all 34 · V1
- The agentic wrapper observatory · V1
- Visual extraction and scoring observation: local MiniCPM-V 4.5 and 4.6 aliases · V3