Ollama Adds Three New Models with /v1/systemone API Support

Three New Models Now Available on Ollama

Ollama, the popular local large language model runner, has announced that three new models are now available on the platform, accessible via the /v1/systemone endpoint. This update gives developers more options for deploying and testing models of different scales locally.

According to the official announcement, the three newly added models are:

• Nimble: A lightweight, fast-response-focused model
• Tev1 4b: A mid-sized model with approximately 4 billion parameters
• Tev1 0.8b: A compact model with approximately 800 million parameters

The naming makes it clear that the Tev1 series offers two variants at different parameter scales, allowing developers to make trade-offs based on their hardware capabilities and performance requirements.

How the Three Models Differ

Nimble: Built for Agile Response

As the name suggests, Nimble is designed with low latency and fast inference in mind. This type of model is typically well-suited for scenarios where response speed matters more than deep reasoning — such as interactive conversations, real-time assistance, and similar use cases.

In the LLM space, "lightweight and fast" is typically achieved through a few technical approaches: reducing total parameter count, applying quantization (compressing model weights from float16 down to int8 or even int4), or pruning the inference path. Nimble's exact parameter count hasn't been disclosed, but based on its positioning, the priority appears to be Time to First Token rather than reasoning depth on complex tasks. In practice, models like this are often used as a "front-end routing layer" — responding to users quickly while optionally handing off more complex requests to a larger model — striking a dynamic balance between user experience and compute consumption.

Tev1 4b and Tev1 0.8b: Covering Different Compute Tiers

The two Tev1 variants form a clear gradient:

• Tev1 4b (~4 billion parameters) strikes a balance between capability and resource usage, making it suitable for local environments with a decent GPU or a relatively powerful CPU.
• Tev1 0.8b (~800 million parameters) is smaller with a lower memory footprint, ideal for resource-constrained devices such as laptops or even edge hardware.

This tiered design is a common pattern in the open-source model ecosystem — offering the same model family at different parameter scales to serve a range of needs, from personal experimentation to lightweight production use.

What the /v1/systemone Endpoint Means

All three models are served through the /v1/systemone endpoint. An API-based interface means developers can access different models using a consistent method, reducing the cost of switching between them. For existing Ollama users, this kind of update allows you to try new models directly within your current workflow without significant code changes.

Ollama's core value proposition has always been "run large models locally with ease," and continuously expanding its model library is key to staying competitive. Every new batch of models gives local inference users more options to compare and combine.

Ollama's API design is highly compatible with OpenAI's /v1/chat/completions specification — a mainstream strategy among local LLM toolchains. This compatibility means developers can use the same client code (such as LangChain, LlamaIndex, or custom scripts) to swap underlying models at any time without modifying business logic. As a dedicated endpoint, /v1/systemone may include specific optimizations for system-level call scenarios, though the official documentation has not yet detailed how it differs behaviorally from the standard chat endpoint. Developers integrating these models are advised to consult Ollama's official API docs to confirm that request formats and parameter fields are fully compatible with their existing workflows.

Practical Value for Developers

For developers and enthusiasts focused on local deployment, this update offers several tangible benefits:

1. More options: From Nimble to the Tev1 series, the new additions cover a range of speed and scale preferences.
2. Easy to try: Pull and run locally via Ollama — no dependency on cloud APIs required.
3. Cost control: Running locally avoids per-call billing, making it ideal for iterative experimentation and privacy-sensitive use cases.

It's worth noting that the official announcement is fairly brief. Details such as training data, benchmark performance, and licensing terms have not been disclosed in the original content. Users interested in adopting these models are encouraged to visit Ollama's official model pages for complete specifications and documentation before evaluating fit for their use case.

Summary

Ollama's three newly added models — Nimble, Tev1 4b, and Tev1 0.8b — further expand the local LLM landscape and are all accessible through the unified /v1/systemone endpoint. While public details remain limited for now, updates like this reflect the vitality of the local AI tooling ecosystem. Developers eager to try them out can head to the Ollama platform to find detailed information on each model.