# Fara 1.5: an open, vision-only browser agent at 4B, 9B, and 27B

> Satyajit Ghana — Head of Engineering @ Inkers Technology
> canonical: https://ai.thesatyajit.com/articles/fara-1-5
> date: 2026-08-03
> tags: agents, computer-use, open-weights, vision-language-models
[Qwen-CUA](/articles/qwen-cua), published earlier today, controls a full desktop from screenshots alone with a 397B-A17B mixture-of-experts model. Fara1.5, from Microsoft Research, does the narrower version of the same job -- browser only, no desktop -- at 4B, 9B, and 27B parameters, and ships all three sizes under the MIT license, weights included. That is two orders of magnitude smaller than Qwen-CUA's backbone, scoped to one surface instead of an entire OS, and open in a way Qwen-CUA specifically isn't: Qwen-CUA's repository is Apache-2.0, but its own README states plainly that "model weights are not included" -- paper and demo only. Fara1.5 publishes the checkpoints. The interesting question isn't whether the biggest Fara model is good. It's what the 4B-to-27B ladder says about how much capability a vision-only agent actually needs.

<Figure
  src="/articles/fara-1-5/fig1.png"
  alt="Diagram of Fara1.5's observe-think-act loop: the model observes three recent screenshots plus conversation history, reasons internally in a thought bubble, then emits one atomic action -- mouse and keyboard, web-specific shortcuts, or context-management meta-actions like ask_user and finish -- before the loop repeats as the browser updates."
  caption="One step of the loop: three recent screenshots and history in, one atomic action out -- no DOM, no accessibility tree (Microsoft Research, Fara1.5 paper, Figure 5)."
/>

## The same narrow interface, a much smaller model

Fara1.5 is a multimodal decoder-only model built on Qwen3.5, at three sizes, all trained the same way. Given a goal, the current screenshot, and the last three steps of history, it reasons in text and then emits exactly one action from a fixed vocabulary: click, type, scroll, drag, `visit_url`, `web_search`, `go_back`, plus meta-actions for longer horizons -- `memorize` to persist a fact past the three-screenshot window, `ask_user` to pause on a critical point, `finish` to stop. Coordinates are predicted directly from pixels, the same design choice Qwen-CUA makes for an entire desktop: skip the DOM and accessibility tree, and bet that whatever a human can operate with a screen and two input devices is a general enough interface. Fara1.5 just makes that bet at a fraction of the parameter count, and only for the browser.

The safety mechanism is worth naming precisely because it's a real constraint, not a suggestion: the model is trained to trigger `ask_user` at eight defined "critical point" types -- across three dimensions (permission granted or not, task fully specified or not, action reversible or not) -- covering things like entering personal information, submitting payment or shipping details, or sending a message on the user's behalf. Microsoft's recommended deployment wrapper, MagenticLite, is sandboxed and pausable specifically so a human is in the loop at those points. `ask_user` and `memorize` are context-management tools in exactly Lilian Weng's sense from [Agent harnesses](/articles/agent-harness) -- deciding what to carry forward and when to hand control back -- just built into the action vocabulary itself rather than the surrounding scaffold.

## Trained on trajectories it generated for itself

Nearly all of Fara1.5's training data comes from FaraGen1.5, its own synthetic-data pipeline. A solver -- GPT-5.4, paired with a user simulator that withholds task details the way a real user would -- attempts tasks in two kinds of environments: the live, open web, and six sandboxed synthetic apps (Mail, Calendar, Stream, ML, Stay, Scheduler) whose functional code was itself generated by a coding agent (GitHub Copilot CLI) rather than scraped or mocked. Every resulting trajectory then has to clear three independent verifiers before it counts as training data.

<FaraGenPipeline />

That data becomes roughly 2 million training samples, 60% still ordinary open-web trajectories, the rest split across synthetic environments, deliberately ambiguous form-filling, grounding, and a small slice of VQA and drag gestures. It's a real answer to the standard complaint about computer-use training data -- human demonstrations are slow and expensive to collect -- but it's worth being precise about what "generated" means here: the solver, the user simulator, and the verifiers are themselves LLM judgments, not ground truth. A verifier checking "did this trajectory ask before an irreversible action" is exactly as reliable as the model doing the checking.

## What the ladder buys

Fara1.5-27B reaches 72.3% on Online-Mind2Web, ahead of Gemini 2.5 Computer Use (57.3%), OpenAI Operator (58.3%), and Yutori Navigator n1 (64.7%) -- three proprietary systems, all evaluated on an independently maintained academic benchmark, not one Microsoft built. That's a genuine result: an open, MIT-licensed family beating closed competitors on a benchmark none of them control. But it only holds at the top of the ladder.

<ScalingCrossover />

WebTailBench v1.5 -- Microsoft's own 609-task eval set, worth flagging as self-authored rather than independent -- shows the same monotonic climb, no crossover to check it against:

<BenchBars
  title="WebTailBench v1.5 (Outcome Success) across the Fara1.5 family"
  unit="%"
  bars={[
    { label: "Fara1.5-4B", value: 27.4 },
    { label: "Fara1.5-9B", value: 32.3 },
    { label: "Fara1.5-27B", value: 40.2, highlight: true },
  ]}
/>

Read against the predecessor, Fara-7B, the jump looks even sharper: Fara1.5-9B improves +29.3 points on Online-Mind2Web, +13.1 on WebVoyager, +8.3 on WebTailBench, +18.1 on ScreenSpot-Pro grounding, +8.9 on OSWorld-G Refined. That comparison is real but not clean -- it conflates a parameter increase (7B to 9B) with a full training-pipeline change (FaraGen1.5 replacing whatever generated the original Fara's data). Stated as a training-pipeline improvement, it overclaims; stated as "the current generation beats the last one," it's exactly as strong as it sounds and no stronger.

<Callout type="note">
The Online-Mind2Web and WebVoyager comparisons against Operator, Gemini 2.5 CU, and Navigator n1 are self-reported by Microsoft on benchmarks those three systems don't control -- a meaningfully better setup than grading your own exam, but still not an independently run leaderboard. No third-party replication of these specific numbers was found for this piece.
</Callout>

## Where the vision-only bet costs something

The model card is direct about the downsides of skipping the DOM: English-only, vulnerable to visual deception and prompt injection embedded in page content, error accumulation over long multi-step trajectories, and explicitly **not suitable** for legal, health, or financial use. None of that is unique to Fara1.5 -- Qwen-CUA's paper documents the same shape of limitation for the same underlying reason -- but a 4B vision-only model has less capacity to notice something is wrong mid-trajectory than a 397B-A17B one, and the model card doesn't pretend otherwise.

## The take

Two orders of magnitude smaller than Qwen-CUA, scoped to a browser instead of a desktop, and shipping actual weights under MIT where Qwen-CUA ships code and a paper but withholds the checkpoints: Fara1.5 is a genuinely different point in the design space, not a smaller copy of the same idea. The headline -- 27B beats three proprietary computer-use agents on a benchmark none of them own -- is real. The more useful reading of the paper is the ladder underneath it: at 4B, Fara1.5 merely ties the weakest of those three baselines; the win only fully arrives at 27B. Vision-only browser control is not a capability a small model gets for free. It's bought, roughly a third of it per step up the ladder, exactly as parameter-scaling laws would predict.

---

*Built on Microsoft Research's [Fara1.5: Scalable Learning Environments for Computer Use Agents](https://arxiv.org/abs/2606.20785) (Awadallah et al., 2026) and the [microsoft/fara](https://github.com/microsoft/fara) repository (MIT license). Figures 5 and 7 are reproduced from the paper for commentary, flattened onto white; the FaraGen1.5 pipeline diagram and scaling-vs-baseline chart are original illustrations of the paper's Figure 2 and Table 3 / Figure 7 data, not measured traces. Benchmark numbers are as reported in the paper.*