Model releases  ·  11 August 2026

Microsoft Fara 1.5: computer-use agents under MIT

A browser agent that looks at screenshots and clicks, released at three sizes under one of the most permissive licences Microsoft has used — and built on a base that is not Microsoft's.

At a glance

Released byMicrosoft Research AI Frontiers
PublishedJuly 2026
Sizes4B, 9B, 27B
Base modelQwen3.5 (Qwen3_5ForConditionalGeneration)
PurposeMultimodal computer-use agent for web browsers
Layers / hidden size32 / 2560 (4B variant)
Context length262,144 tokens
Vocabulary248,320 tokens
LicenceMIT
Weight size9.11 GB for the 4B variant
Hosted atmicrosoft on Hugging Face

What it is

Fara 1.5 is a computer-use agent: it observes a web browser through screenshots and acts by emitting structured tool calls — click, type, scroll, visit URL, web search — to complete a task end to end. That is a narrower and more concrete job than general assistance, and it is the reason the models are multimodal.

Microsoft positions the 4B as viable on a single accelerator with room for screenshot history, which matters for an agent that has to remember what the page looked like several steps ago.

Built on Qwen3.5

The configuration is explicit: Qwen3_5ForConditionalGeneration, with a 248,320-token vocabulary and a 262,144-token context — the Qwen3.5 signature. Microsoft Research built its computer-use agent on Alibaba's open base rather than on Phi, its own small-model line.

That is a reasonable read on where the Phi line stands: the most recent Phi releases predate this, and Microsoft's newer open work is appearing under the Fara and Mage names.

The licence

MIT across all three sizes, with no acceptable-use appendix. For a computer-use agent — a category where vendors often attach usage restrictions — that is notable.

Runtime support

RuntimeSupportedNotes
transformersYesReference implementation from Microsoft.
llama.cppPartialThe underlying Qwen3.5 architecture is supported, but the screenshot-driven agent loop is not something a plain text runtime provides.
MLXPartialCommunity conversions only.
Vision inputYesCentral to the design; the model reads browser screenshots.

What we don't know yet

Sources

Frequently asked questions

What is Fara 1.5?

A family of multimodal computer-use agent models from Microsoft Research that drive a web browser by reading screenshots and emitting structured tool calls. Released at 4B, 9B and 27B.

What licence is Fara 1.5 under?

MIT, across all three sizes, with no additional acceptable-use policy.

What is Fara 1.5 built on?

Qwen3.5. The configuration declares the Qwen3.5 architecture, vocabulary and context length rather than a Microsoft base such as Phi.

How large is Fara1.5-4B?

9.11 GB in the published safetensors form.

Can I run Fara 1.5 in llama.cpp?

The underlying architecture is supported, but Fara is an agent that consumes screenshots and emits tool calls — a plain text inference runtime does not provide that loop.

OnDevice LLM is a private AI assistant that runs entirely on your iPhone — no account, no cloud, and nothing you type leaves the device.

Published 11 August 2026