Head to head · Browser automation · October 2026 research run
Fara1.5-4B vs Fara1.5-9B
Fara1.5-4B and Fara1.5-9B score within a point of each other on agent readiness, 61 (C) and 61 (C). Both do browser automation.
Best desktop and mobile app automation for AI agents · All 55 desktop comparisons
Which one, for what
Best for Developers building a supervised web-browsing agent who want open weights and clear safety behaviour
Ahead on
- Reach, 75 against 70
Watch for
Screenshots only, so a page can mislead or instruct it
Best for Developers building a supervised web-browsing agent who want open weights and clear safety behaviour
No category where it leads by five points or more, and no fact that sets it apart.
Watch for
Screenshots only, so a page can mislead or instruct it
Why the edge
Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.
Reach Fara1.5-4B +5
Fara1.5-4B 75
Memory 30, the only build from Microsoft is the BF16 safetensors, 9.08 GB, and 9.08 GB × 1.2 = 10.9 GB, within 16 GB. bartowski's GGUF isn't read for this line. Runtimes 15, the architecture is Qwen3_5ForConditionalGeneration, which vLLM 0.33.0, SGLang 0.5.23 and Transformers 5.22.2 run from the safetensors files, three at 5 each. Builds 0, Microsoft publishes the BF16 weights only. Terms 15, MIT in the card's metadata and its licence section, an OSI licence with nothing added. Download 10, the Hub API reports gated false on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, the card links an Azure Foundry page for this model, Microsoft's own platform. 30 + 15 + 0 + 15 + 10 + 5 = 75.
Fara1.5-9B 70
Memory 25, the only build from Microsoft is the BF16 safetensors, 18.82 GB, and 18.82 GB × 1.2 = 22.6 GB, within 24 GB. bartowski's GGUF isn't read for this line. Runtimes 15, the architecture is Qwen3_5ForConditionalGeneration, which vLLM 0.33.0, SGLang 0.5.23 and Transformers 5.22.2 run from the safetensors files, three at 5 each. Builds 0, Microsoft publishes the BF16 weights only. Terms 15, MIT in the card's metadata and its licence section, an OSI licence with nothing added. Download 10, the Hub API reports gated false on 10 November 2026, so the files download with no account and nothing to accept. Hosted 5, the card links an Azure Foundry page for this model, Microsoft's own platform. 25 + 15 + 0 + 15 + 10 + 5 = 70.
Score by category
| Category | Weight this run | Fara1.5-4B | Fara1.5-9B | Edge |
|---|---|---|---|---|
| Capability | 35%pending | pending | pending | not scored in this run |
| Reach | 25%25 | 75 | 70 | Fara1.5-4B +5 |
| Risk | 20%20 | 75 | 75 | even |
| Agent interface | 10%10 | 80 | 80 | even |
| Documentation and upkeep | 10%10 | 60 | 60 | even |
| Negative events | ≤15 | 0 | 0 | |
| Total | C 61/100 | C 61/100 |
Facts side by side
| Fact | Fara1.5-4B | Fara1.5-9B |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Microsoft | Microsoft |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | MIT | MIT |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-07-27 | 2026-07-27 |
| Terms last updated | no document linked | no document linked |
| Privacy policy last updated | no document linked | no document linked |
Verdicts
Fara1.5-4B
A browser agent model with unusually careful documentation: what it does, what it's trained to stop at, what it gets wrong and how to box it in, under MIT. Treat it as a component for Microsoft's harness, in a sandbox, with a person on the irreversible steps. Safety testing is described but not published, and it's English only.
Fara1.5-9B
A browser agent model with unusually careful documentation: what it does, what it's trained to stop at, what it gets wrong and how to box it in, under MIT. Treat it as a component for Microsoft's harness, in a sandbox, with a person on the irreversible steps. Safety testing is described but not published, and it's English only.
Before you call either
Fara1.5-4B
- Run it inside Microsoft's harness (the fara CLI or MagenticLite), in a sandboxed browser with an allow-list. The card says not to give it open access on a machine with anything sensitive
- Send the card's system prompt word for word. The model was trained against it
- Match a 1440×900 screen for reliable clicks
- Expect it to stop and ask before purchases, sign-ins, messages or personal data unless the user authorised them up front. Pass the question to the user; don't answer for them
- Treat page text as untrusted. The model can be redirected by instructions on a page
- It emits one action per turn at temperature 0, and ends with
terminate
Fara1.5-9B
- Run it inside Microsoft's harness (the fara CLI or MagenticLite), in a sandboxed browser with an allow-list. The card says not to give it open access on a machine with anything sensitive
- Send the card's system prompt word for word. The model was trained against it
- Match a 1440×900 screen for reliable clicks
- Expect it to stop and ask before purchases, sign-ins, messages or personal data unless the user authorised them up front. Pass the question to the user; don't answer for them
- Treat page text as untrusted. The model can be redirected by instructions on a page
- It emits one action per turn at temperature 0, and ends with
terminate
Questions
Which is better for AI agents, Fara1.5-4B or Fara1.5-9B?
Fara1.5-4B and Fara1.5-9B score within a point of each other on agent readiness, 61 (C) and 61 (C).
Other comparisons with Fara1.5-4B or Fara1.5-9B
Machine-readable
- This page as Markdown
/compare/fara1-5-4b-vs-fara1-5-9b.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/fara1-5-4b.json·/api/v1/tools/fara1-5-9b.json - From a terminal
anchor compare fara1-5-4b fara1-5-9b(the CLI) - Over MCP
compare_tools {"a": "fara1-5-4b", "b": "fara1-5-9b"}at/mcp, no key