{ "data": { "a": { "slug": "fara1-5-27b", "name": "Fara1.5-27B", "vendor": "Microsoft", "vendorUrl": "https://github.com/microsoft/fara", "kind": "weights", "category": "desktop-automation", "summary": "Fara1.5-27B is Microsoft's open-weight computer use model for web browsers, 27B parameters fine-tuned from Qwen3.5-27B. It reads screenshots and returns one click, keystroke or navigation at a time, pausing before irreversible steps. MIT, English only.", "url": "https://www.anchorterminal.com/tools/fara1-5-27b", "markdownUrl": "https://www.anchorterminal.com/tools/fara1-5-27b.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/fara1-5-27b.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/fara1-5-27b.json", "dns": "fara1-5-27b.grade.anchorterminal.com", "repo": "https://huggingface.co/microsoft/Fara1.5-27B", "license": "MIT", "transports": [], "packages": [ { "registry": "huggingface", "name": "microsoft/Fara1.5-27B" } ], "auth": "none", "authNotes": "The weights download from Hugging Face with no account, token or terms to accept. The repository is not gated (checked 2026-11-10).", "pricing": "free", "pricingNotes": "Free to download and use under MIT. Running it yourself takes 54.71 GB in BF16, the only build Microsoft publishes. The card says to run it inside Microsoft's own harness, the fara CLI or MagenticLite (checked 2026-11-10).", "priceSummary": "Free", "where": "local", "x402": { "level": "no", "evidence": "Weights you download and run yourself, so there is nothing to pay per call and no payment protocol (checked 2026-11-10).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-11-10" }, "docsUrl": "https://github.com/microsoft/fara", "capabilities": [ "browser.control", "desktop.control", "model.open-weights" ], "tags": [ "open-weights", "model", "computer-use", "browser-agent", "gui-agent", "vision", "tool-calling", "english", "mit", "safetensors", "self-hosted" ], "lastRelease": "2026-07-27", "graded": true, "anchor": { "graded": true, "score": 61, "grade": "C", "agentReady": false, "rank": 0, "ranked": false, "notRankedWhy": "An open-weight model, graded by the open-weight checklist and ranked among open-weight models, not against tools", "rankOf": 1518, "categoryRank": 4, "methodology": "0.7", "run": "2026-10-01", "scores": { "interface": 80, "reach": 55, "risk": 75, "upkeep": 65 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "low", "date": "2026-11-10" }, "negative": 0, "verdict": "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.", "bestFor": "Developers building a supervised web-browsing agent who want open weights and clear safety behaviour", "strengths": [ "Trained to pause before purchases, sign-ins, messages and entering personal data", "A complete card, with limitations, data make-up and a cut-off", "MIT, ungated", "Every action is a JSON call against a documented schema", "Runs in released vLLM, SGLang and Transformers with no repository code" ], "weaknesses": [ "Screenshots only, so a page can mislead or instruct it", "Safety red-teaming is described without results", "English only", "One build, in BF16, and no GGUF from Microsoft", "The card says not to use it in production without more testing and safeguards", "The repository's history was squashed" ], "agentNotes": [ "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`" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "low", "grade": "C", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 61 } ], "editorialScores": { "interface": 80, "reach": 55, "risk": 75, "upkeep": 65 }, "provenanceScore": 67, "checked": { "categories": { "interface": "2026-11-10", "reach": "2026-11-10", "risk": "2026-11-10", "upkeep": "2026-11-10" }, "last": "2026-11-10", "oldest": "2026-11-10", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "pip install \"vllm\u003e=0.33.0\"\nvllm serve microsoft/Fara1.5-27B --dtype bfloat16 --max-model-len 262144 --limit-mm-per-prompt image=10\n# then drive it with Microsoft's harness: github.com/microsoft/fara, or MagenticLite (github.com/microsoft/magentic-ui)", "http": "# vLLM serves an OpenAI-compatible API at http://localhost:8000/v1. Send the card's system prompt word for word,\n# the goal and the latest screenshot as an image part, at temperature 0. The reply ends in one tool_call block" }, "letme": { "capability": "https://letme.dev/browser.control", "tool": "https://letme.dev/fara1-5-27b" }, "sameCompany": [ "azure-openai", "azure-foundry-fine-tuning", "azure-ai-content-safety", "azure-speech-to-text", "azure-text-to-speech", "microsoft-agent-framework", "microsoft-execution-containers", "microsoft-entra-agent-id", "azure-key-vault", "bing-grounding", "azure-document-intelligence", "azure-devops-mcp", "microsoft-learn-mcp", "tokeneconomizer", "playwright-mcp", "azure-mcp", "azure-maps", "azure-translator", "microsoft-graph-calendar", "azure-blob-storage", "onedrive-sharepoint", "microsoft-teams", "dynamics-365-sales", "copilot-studio-mcp", "power-automate", "foundry-local", "microsoft-decision-1", "microsoft-advertising-api", "microsoft-excel-graph", "outlook-mail-graph", "microsoft-graph-people", "fara1-5-4b", "fara1-5-9b", "mage-vl-cowork" ], "weights": { "trainedBy": "Microsoft Research AI Frontiers", "builtOn": "Qwen3.5-27B, fine-tuned, per the card and the metadata", "params": "27B", "context": 262144, "goodFor": "Driving a web browser from screenshots inside a sandbox: forms, bookings, shopping and look-ups, with a person approving the irreversible steps. Fits a 96 GB machine in BF16, or a 24 GB GPU with a third party's 4-bit GGUF.", "runtimes": [ "vLLM", "SGLang", "Transformers", "llama.cpp" ], "builds": [ { "name": "bartowski Fara1.5-27B-GGUF (Q4_K_M)", "url": "https://huggingface.co/bartowski/Fara1.5-27B-GGUF", "who": "bartowski", "precision": "q4_k_m", "sizeGB": 18.46, "memoryGB": 24, "format": "gguf", "runsOn": [ "llama.cpp" ], "note": "From bartowski, a third party's conversion and not Microsoft's. 17.53 GB of Q4_K_M weights plus the 0.93 GB F16 projector, from the Hub's model information on 10 November 2026. Reach doesn't count it" }, { "name": "bartowski Fara1.5-27B-GGUF (Q8_0)", "url": "https://huggingface.co/bartowski/Fara1.5-27B-GGUF", "who": "bartowski", "precision": "q8_0", "sizeGB": 29.6, "memoryGB": 48, "format": "gguf", "runsOn": [ "llama.cpp" ], "note": "From bartowski, a third party's conversion and not Microsoft's. 28.67 GB of Q8_0 weights plus the 0.93 GB F16 projector, from the Hub's model information on 10 November 2026. Reach doesn't count it" }, { "name": "Fara1.5-27B (BF16 safetensors)", "url": "https://huggingface.co/microsoft/Fara1.5-27B", "who": "publisher", "precision": "bf16", "sizeGB": 54.71, "memoryGB": 96, "format": "safetensors", "runsOn": [ "vLLM", "SGLang", "Transformers" ], "note": "The only build Microsoft publishes" } ], "run": "vllm serve microsoft/Fara1.5-27B --dtype bfloat16 --max-model-len 262144 --limit-mm-per-prompt image=10", "runBuild": "Fara1.5-27B (BF16 safetensors)", "checked": "2026-11-10", "minMemoryGB": 24 }, "family": { "lab": "Microsoft", "family": "Fara", "url": "https://www.anchorterminal.com/open-weights/#microsoft", "others": [ { "name": "Fara1.5-9B", "url": "https://huggingface.co/microsoft/Fara1.5-9B", "params": "9B (9,409,813,744 by the Hub's safetensors count)", "listing": "https://www.anchorterminal.com/tools/fara1-5-9b" }, { "name": "Fara1.5-4B", "url": "https://huggingface.co/microsoft/Fara1.5-4B", "params": "4.5B (4,539,265,536 by the Hub's safetensors count)", "listing": "https://www.anchorterminal.com/tools/fara1-5-4b" } ] }, "area": "developer", "provenance": { "legalEntity": "Microsoft", "domain": "microsoft.com", "domainRegistered": "", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "https://huggingface.co/microsoft/Fara1.5-27B/commits/main", "securityTxt": "expired", "checked": "2026-11-10", "notes": [ "The files sit in Microsoft's microsoft organisation on Hugging Face, and the card names Microsoft Research AI Frontiers as developer. Domain and security.txt are carried from our other Microsoft listings.", "Nothing is hosted under this listing, so MIT stands in for terms of service and no privacy policy is scored.", "The changelog is the commit history. One commit, a squash of the branch on 27 July 2026, so earlier history isn't visible.", "The card names Microsoft Ireland Operations Limited as authorised representative for EU AI Act requests." ], "score": 67 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/fara1-5-27b.json" }, "answer": "Fara1.5-27B and Fara1.5-4B score within a point of each other on agent readiness, 61 (C) and 61 (C). Fara1.5-4B leads on reach.", "b": { "slug": "fara1-5-4b", "name": "Fara1.5-4B", "vendor": "Microsoft", "vendorUrl": "https://github.com/microsoft/fara", "kind": "weights", "category": "desktop-automation", "summary": "Fara1.5-4B is Microsoft's open-weight computer use model for web browsers, 4.5B parameters fine-tuned from Qwen3.5-4B. It reads screenshots and returns one click, keystroke or navigation at a time, pausing before irreversible steps. MIT, English only.", "url": "https://www.anchorterminal.com/tools/fara1-5-4b", "markdownUrl": "https://www.anchorterminal.com/tools/fara1-5-4b.md", "slimMarkdownUrl": "https://www.anchorterminal.com/tools/fara1-5-4b.min.md", "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/fara1-5-4b.json", "dns": "fara1-5-4b.grade.anchorterminal.com", "repo": "https://huggingface.co/microsoft/Fara1.5-4B", "license": "MIT", "transports": [], "packages": [ { "registry": "huggingface", "name": "microsoft/Fara1.5-4B" } ], "auth": "none", "authNotes": "The weights download from Hugging Face with no account, token or terms to accept. The repository is not gated (checked 2026-11-10).", "pricing": "free", "pricingNotes": "Free to download and use under MIT. Running it yourself takes 9.08 GB in BF16, the only build Microsoft publishes. The card says to run it inside Microsoft's own harness, the fara CLI or MagenticLite (checked 2026-11-10).", "priceSummary": "Free", "where": "local", "x402": { "level": "no", "evidence": "Weights you download and run yourself, so there is nothing to pay per call and no payment protocol (checked 2026-11-10).", "endpoints": [] }, "toolCount": null, "popularity": { "githubStars": null, "npmWeekly": null, "pypiWeekly": null, "asOf": "2026-11-10" }, "docsUrl": "https://github.com/microsoft/fara", "capabilities": [ "browser.control", "desktop.control", "model.open-weights" ], "tags": [ "open-weights", "model", "computer-use", "browser-agent", "gui-agent", "vision", "tool-calling", "english", "mit", "safetensors", "self-hosted" ], "lastRelease": "2026-07-27", "graded": true, "anchor": { "graded": true, "score": 61, "grade": "C", "agentReady": false, "rank": 0, "ranked": false, "notRankedWhy": "An open-weight model, graded by the open-weight checklist and ranked among open-weight models, not against tools", "rankOf": 1518, "categoryRank": 5, "methodology": "0.7", "run": "2026-10-01", "scores": { "interface": 80, "reach": 75, "risk": 75, "upkeep": 60 }, "pending": [ "performance", "tasks" ], "assessment": { "confidence": "low", "date": "2026-11-10" }, "negative": 0, "verdict": "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.", "bestFor": "Developers building a supervised web-browsing agent who want open weights and clear safety behaviour", "strengths": [ "Trained to pause before purchases, sign-ins, messages and entering personal data", "A complete card, with limitations, data make-up and a cut-off", "MIT, ungated", "Every action is a JSON call against a documented schema", "Runs in released vLLM, SGLang and Transformers with no repository code" ], "weaknesses": [ "Screenshots only, so a page can mislead or instruct it", "Safety red-teaming is described without results", "English only", "One build, in BF16, and no GGUF from Microsoft", "The card says not to use it in production without more testing and safeguards", "The repository's history was squashed" ], "agentNotes": [ "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`" ], "metrics": { "kind": "local", "measured": false }, "reviewCount": 0, "avgRating": 0, "history": [ { "basis": "public evidence", "confidence": "low", "grade": "C", "methodology": "0.7", "pending": [ "performance", "tasks" ], "run": "2026-10-01", "runLabel": "October 2026 research run", "score": 61 } ], "editorialScores": { "interface": 80, "reach": 75, "risk": 75, "upkeep": 60 }, "provenanceScore": 67, "checked": { "categories": { "interface": "2026-11-10", "reach": "2026-11-10", "risk": "2026-11-10", "upkeep": "2026-11-10" }, "last": "2026-11-10", "oldest": "2026-11-10", "stale": false, "staleAfterDays": 90 } }, "connect": { "install": "pip install \"vllm\u003e=0.33.0\"\nvllm serve microsoft/Fara1.5-4B --dtype bfloat16 --max-model-len 262144 --limit-mm-per-prompt image=10\n# then drive it with Microsoft's harness: github.com/microsoft/fara, or MagenticLite (github.com/microsoft/magentic-ui)", "http": "# vLLM serves an OpenAI-compatible API at http://localhost:8000/v1. Send the card's system prompt word for word,\n# the goal and the latest screenshot as an image part, at temperature 0. The reply ends in one tool_call block" }, "letme": { "capability": "https://letme.dev/browser.control", "tool": "https://letme.dev/fara1-5-4b" }, "sameCompany": [ "azure-openai", "azure-foundry-fine-tuning", "azure-ai-content-safety", "azure-speech-to-text", "azure-text-to-speech", "microsoft-agent-framework", "microsoft-execution-containers", "microsoft-entra-agent-id", "azure-key-vault", "bing-grounding", "azure-document-intelligence", "azure-devops-mcp", "microsoft-learn-mcp", "tokeneconomizer", "playwright-mcp", "azure-mcp", "azure-maps", "azure-translator", "microsoft-graph-calendar", "azure-blob-storage", "onedrive-sharepoint", "microsoft-teams", "dynamics-365-sales", "copilot-studio-mcp", "power-automate", "foundry-local", "microsoft-decision-1", "microsoft-advertising-api", "microsoft-excel-graph", "outlook-mail-graph", "microsoft-graph-people", "fara1-5-27b", "fara1-5-9b", "mage-vl-cowork" ], "weights": { "trainedBy": "Microsoft Research AI Frontiers", "builtOn": "Qwen3.5-4B, fine-tuned, per the card and the metadata", "params": "4.5B", "context": 262144, "goodFor": "Driving a web browser from screenshots inside a sandbox: forms, bookings, shopping and look-ups, with a person approving the irreversible steps. Fits a 12 GB GPU in BF16.", "runtimes": [ "vLLM", "SGLang", "Transformers", "llama.cpp" ], "builds": [ { "name": "bartowski Fara1.5-4B-GGUF (Q4_K_M)", "url": "https://huggingface.co/bartowski/Fara1.5-4B-GGUF", "who": "bartowski", "precision": "q4_k_m", "sizeGB": 3.55, "memoryGB": 8, "format": "gguf", "runsOn": [ "llama.cpp" ], "note": "From bartowski, a third party's conversion and not Microsoft's. 2.88 GB of Q4_K_M weights plus the 0.67 GB F16 projector, from the Hub's model information on 10 November 2026. Reach doesn't count it" }, { "name": "bartowski Fara1.5-4B-GGUF (Q8_0)", "url": "https://huggingface.co/bartowski/Fara1.5-4B-GGUF", "who": "bartowski", "precision": "q8_0", "sizeGB": 5.16, "memoryGB": 8, "format": "gguf", "runsOn": [ "llama.cpp" ], "note": "From bartowski, a third party's conversion and not Microsoft's. 4.49 GB of Q8_0 weights plus the 0.67 GB F16 projector, from the Hub's model information on 10 November 2026. Reach doesn't count it" }, { "name": "Fara1.5-4B (BF16 safetensors)", "url": "https://huggingface.co/microsoft/Fara1.5-4B", "who": "publisher", "precision": "bf16", "sizeGB": 9.08, "memoryGB": 12, "format": "safetensors", "runsOn": [ "vLLM", "SGLang", "Transformers" ], "note": "The only build Microsoft publishes" } ], "run": "vllm serve microsoft/Fara1.5-4B --dtype bfloat16 --max-model-len 262144 --limit-mm-per-prompt image=10", "runBuild": "Fara1.5-4B (BF16 safetensors)", "checked": "2026-11-10", "minMemoryGB": 8 }, "family": { "lab": "Microsoft", "family": "Fara", "url": "https://www.anchorterminal.com/open-weights/#microsoft", "others": [ { "name": "Fara1.5-27B", "url": "https://huggingface.co/microsoft/Fara1.5-27B", "params": "27B (27,356,728,560 by the Hub's safetensors count)", "listing": "https://www.anchorterminal.com/tools/fara1-5-27b" }, { "name": "Fara1.5-9B", "url": "https://huggingface.co/microsoft/Fara1.5-9B", "params": "9B (9,409,813,744 by the Hub's safetensors count)", "listing": "https://www.anchorterminal.com/tools/fara1-5-9b" } ] }, "area": "developer", "provenance": { "legalEntity": "Microsoft", "domain": "microsoft.com", "domainRegistered": "", "endpointOnVendorDomain": null, "terms": "", "privacy": "", "statusPage": "", "changelog": "https://huggingface.co/microsoft/Fara1.5-4B/commits/main", "securityTxt": "expired", "checked": "2026-11-10", "notes": [ "The files sit in Microsoft's microsoft organisation on Hugging Face, and the card names Microsoft Research AI Frontiers as developer. Domain and security.txt are carried from our other Microsoft listings.", "Nothing is hosted under this listing, so MIT stands in for terms of service and no privacy policy is scored.", "The changelog is the commit history. Two commits, a squash of the branch on 21 July 2026 and a card update on 27 July, so earlier history isn't visible.", "The card names Microsoft Ireland Operations Limited as authorised representative for EU AI Act requests." ], "score": 67 }, "pageJsonUrl": "https://www.anchorterminal.com/tools/fara1-5-4b.json" }, "edges": [ { "by": 20, "category": "reach", "leads": "fara1-5-4b", "name": "Reach", "reasons": { "a": "Memory 10, the only build from Microsoft is the BF16 safetensors, 54.71 GB, and 54.71 GB × 1.2 = 65.7 GB, which fits the reference machine but not 64 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, and Featherless is live for it on Hugging Face. 10 + 15 + 0 + 15 + 10 + 5 = 55.", "b": "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." }, "scores": { "a": 55, "b": 75 } }, { "by": 5, "category": "upkeep", "leads": "fara1-5-27b", "name": "Documentation and upkeep", "reasons": { "a": "Model card 20, Intended use, out-of-scope use and known limitations, the architecture, the parameters, the context, the language (English only) and the data cut-off (20 March 2026) are all on the card. Training data 15, A data section that names the kinds of source and the mix, the scale, the cut-off (last acquisition 20 March 2026) and the filtering (a verifier keeps only successful trajectories). Last change 10, the last commit, on 27 July 2026, is 106 days before this check and within 180. Responsiveness 5, Of the 5 discussions that aren't pull requests (22 to 29 July 2026), 1 has a reply from a microsoft member, which is a few, counted on 10 November by author flags and dates with a script, text not read. Version policy 5, the card says future updates will ship as separately versioned models with their own cards, and the earlier Fara model stays in its own repository. File health 10, Config, template and card agree on the architecture and the 262,144-token context. 20 + 15 + 10 + 5 + 5 + 10 = 65.", "b": "Model card 20, Intended use, out-of-scope use and known limitations, the architecture, the parameters, the context, the language (English only) and the data cut-off (20 March 2026) are all on the card. Training data 15, A data section that names the kinds of source and the mix, the scale, the cut-off (last acquisition 20 March 2026) and the filtering (a verifier keeps only successful trajectories). Last change 10, the last commit, on 27 July 2026, is 106 days before this check and within 180. Responsiveness 0, the one discussion that isn't a pull request (22 July 2026) has no reply from a microsoft member, counted on 10 November by author flags and dates with a script, text not read. Version policy 5, the card says future updates will ship as separately versioned models with their own cards, and the earlier Fara model stays in its own repository. File health 10, Config, template and card agree on the architecture and the 262,144-token context. 20 + 15 + 10 + 0 + 5 + 10 = 60." }, "scores": { "a": 65, "b": 60 } } ], "facts": [ { "a": "Open-weight model", "b": "Open-weight model", "name": "Kind" }, { "a": "Microsoft", "b": "Microsoft", "name": "Vendor" }, { "a": "no (local only)", "b": "no (local only)", "name": "Hosted endpoint" }, { "a": "", "b": "", "name": "Transports" }, { "a": "None", "b": "None", "name": "Auth" }, { "a": "Free", "b": "Free", "name": "Pricing" }, { "a": "no", "b": "no", "name": "x402" }, { "a": "MIT", "b": "MIT", "name": "Licence" }, { "a": "no", "b": "no", "name": "Read-only variant documented" }, { "a": "no", "b": "no", "name": "llms.txt" }, { "a": "2026-07-27", "b": "2026-07-27", "name": "Last release" }, { "a": "no document linked", "b": "no document linked", "name": "Terms last updated" }, { "a": "no document linked", "b": "no document linked", "name": "Privacy policy last updated" } ], "faq": [ { "answer": "Fara1.5-27B and Fara1.5-4B score within a point of each other on agent readiness, 61 (C) and 61 (C). Fara1.5-4B leads on reach.", "question": "Which is better for AI agents, Fara1.5-27B or Fara1.5-4B?" } ], "goodFor": [ { "aheadOn": [ "Documentation and upkeep, 65 against 60" ], "also": null, "goodFor": "Developers building a supervised web-browsing agent who want open weights and clear safety behaviour", "slug": "fara1-5-27b", "watchFor": "Screenshots only, so a page can mislead or instruct it" }, { "aheadOn": [ "Reach, 75 against 55" ], "also": null, "goodFor": "Developers building a supervised web-browsing agent who want open weights and clear safety behaviour", "slug": "fara1-5-4b", "watchFor": "Screenshots only, so a page can mislead or instruct it" } ], "history": null, "job": { "capability": "browser.control", "name": "Browser automation" }, "others": [ { "json": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-9b.json", "title": "Fara1.5-27B vs Fara1.5-9B", "url": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-9b" }, { "json": "https://www.anchorterminal.com/compare/fara1-5-4b-vs-fara1-5-9b.json", "title": "Fara1.5-4B vs Fara1.5-9B", "url": "https://www.anchorterminal.com/compare/fara1-5-4b-vs-fara1-5-9b" }, { "json": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-mage-vl-cowork.json", "title": "Fara1.5-27B vs Mage-VL-Cowork", "url": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-mage-vl-cowork" }, { "json": "https://www.anchorterminal.com/compare/fara1-5-4b-vs-mage-vl-cowork.json", "title": "Fara1.5-4B vs Mage-VL-Cowork", "url": "https://www.anchorterminal.com/compare/fara1-5-4b-vs-mage-vl-cowork" } ], "panel": null, "scores": [ { "key": "capability", "name": "Capability", "pending": true, "weight": 35 }, { "by": 20, "edge": "fara1-5-4b", "fara1-5-27b": 55, "fara1-5-4b": 75, "key": "reach", "name": "Reach", "weight": 25 }, { "by": 0, "edge": "", "fara1-5-27b": 75, "fara1-5-4b": 75, "key": "risk", "name": "Risk", "weight": 20 }, { "by": 0, "edge": "", "fara1-5-27b": 80, "fara1-5-4b": 80, "key": "interface", "name": "Agent interface", "weight": 10 }, { "by": 5, "edge": "fara1-5-27b", "fara1-5-27b": 65, "fara1-5-4b": 60, "key": "upkeep", "name": "Documentation and upkeep", "weight": 10 } ], "summary": "Fara1.5-27B and Fara1.5-4B score within a point of each other on agent readiness, 61 (C) and 61 (C). Fara1.5-4B leads on reach. Both do browser automation.", "tasks": null, "tutorials": [], "verdicts": { "fara1-5-27b": "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.", "fara1-5-4b": "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." } }, "kind": "anchor.page", "links": { "api": "https://www.anchorterminal.com/api/v1/index.json", "html": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b", "json": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b.json", "llms": "https://www.anchorterminal.com/llms.txt", "markdown": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b.md", "mcp": "https://www.anchorterminal.com/mcp", "slim": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b.min.md" }, "markdown": "Fara1.5-27B and Fara1.5-4B score within a point of each other on agent readiness, 61 (C) and 61 (C). Fara1.5-4B leads on reach. Both do browser automation.\n\n- Fara1.5-27B: grade C, 61/100, rank #62 of 160 open-weight models. Markdown https://www.anchorterminal.com/tools/fara1-5-27b.md · JSON https://www.anchorterminal.com/api/v1/tools/fara1-5-27b.json\n- Fara1.5-4B: grade C, 61/100, rank #63 of 160 open-weight models. Markdown https://www.anchorterminal.com/tools/fara1-5-4b.md · JSON https://www.anchorterminal.com/api/v1/tools/fara1-5-4b.json\n- Best desktop and mobile app automation for AI agents: https://www.anchorterminal.com/best/desktop-automation/index.md\n- All 55 desktop comparisons: https://www.anchorterminal.com/compare/desktop-automation/index.md\n\n## Which one, for what\n\n### Fara1.5-27B (C)\n\nGood for: Developers building a supervised web-browsing agent who want open weights and clear safety behaviour\n\nAhead on:\n- Documentation and upkeep, 65 against 60\n\nWatch for: Screenshots only, so a page can mislead or instruct it\n\n### Fara1.5-4B (C)\n\nGood for: Developers building a supervised web-browsing agent who want open weights and clear safety behaviour\n\nAhead on:\n- Reach, 75 against 55\n\nWatch for: Screenshots only, so a page can mislead or instruct it\n\n\n## Why the edge\n\n### Reach: Fara1.5-4B +20\n\n- Fara1.5-27B (55): Memory 10, the only build from Microsoft is the BF16 safetensors, 54.71 GB, and 54.71 GB × 1.2 = 65.7 GB, which fits the reference machine but not 64 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, and Featherless is live for it on Hugging Face. 10 + 15 + 0 + 15 + 10 + 5 = 55.\n- 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.\n\n### Documentation and upkeep: Fara1.5-27B +5\n\n- Fara1.5-27B (65): Model card 20, Intended use, out-of-scope use and known limitations, the architecture, the parameters, the context, the language (English only) and the data cut-off (20 March 2026) are all on the card. Training data 15, A data section that names the kinds of source and the mix, the scale, the cut-off (last acquisition 20 March 2026) and the filtering (a verifier keeps only successful trajectories). Last change 10, the last commit, on 27 July 2026, is 106 days before this check and within 180. Responsiveness 5, Of the 5 discussions that aren't pull requests (22 to 29 July 2026), 1 has a reply from a microsoft member, which is a few, counted on 10 November by author flags and dates with a script, text not read. Version policy 5, the card says future updates will ship as separately versioned models with their own cards, and the earlier Fara model stays in its own repository. File health 10, Config, template and card agree on the architecture and the 262,144-token context. 20 + 15 + 10 + 5 + 5 + 10 = 65.\n- Fara1.5-4B (60): Model card 20, Intended use, out-of-scope use and known limitations, the architecture, the parameters, the context, the language (English only) and the data cut-off (20 March 2026) are all on the card. Training data 15, A data section that names the kinds of source and the mix, the scale, the cut-off (last acquisition 20 March 2026) and the filtering (a verifier keeps only successful trajectories). Last change 10, the last commit, on 27 July 2026, is 106 days before this check and within 180. Responsiveness 0, the one discussion that isn't a pull request (22 July 2026) has no reply from a microsoft member, counted on 10 November by author flags and dates with a script, text not read. Version policy 5, the card says future updates will ship as separately versioned models with their own cards, and the earlier Fara model stays in its own repository. File health 10, Config, template and card agree on the architecture and the 262,144-token context. 20 + 15 + 10 + 0 + 5 + 10 = 60.\n\n## Score by category\n\n| Category | Weight | Fara1.5-27B | Fara1.5-4B | Edge |\n| --- | --- | --- | --- | --- |\n| Capability | 35%, pending | pending | pending | not scored in this run |\n| Reach | 25% (25 this run) | 55 | 75 | Fara1.5-4B +20 |\n| Risk | 20% (20 this run) | 75 | 75 | even |\n| Agent interface | 10% (10 this run) | 80 | 80 | even |\n| Documentation and upkeep | 10% (10 this run) | 65 | 60 | Fara1.5-27B +5 |\n| Negative events | ≤15 | 0 | 0 | |\n| **Total** | | **61 · C** | **61 · C** | |\n\n## Facts side by side\n\n| Fact | Fara1.5-27B | Fara1.5-4B |\n| --- | --- | --- |\n| Kind | Open-weight model | Open-weight model |\n| Vendor | Microsoft | Microsoft |\n| Hosted endpoint | no (local only) | no (local only) |\n| Transports | | |\n| Auth | None | None |\n| Pricing | Free | Free |\n| x402 | no | no |\n| Licence | MIT | MIT |\n| Read-only variant documented | no | no |\n| llms.txt | no | no |\n| Last release | 2026-07-27 | 2026-07-27 |\n| Terms last updated | no document linked | no document linked |\n| Privacy policy last updated | no document linked | no document linked |\n\n## Verdicts\n\n**Fara1.5-27B.** 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.\n\n**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.\n\n## Before you call either\n\n### Fara1.5-27B\n\n1. 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\n2. Send the card's system prompt word for word. The model was trained against it\n3. Match a 1440×900 screen for reliable clicks\n4. 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\n5. Treat page text as untrusted. The model can be redirected by instructions on a page\n6. It emits one action per turn at temperature 0, and ends with `terminate`\n\n### Fara1.5-4B\n\n1. 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\n2. Send the card's system prompt word for word. The model was trained against it\n3. Match a 1440×900 screen for reliable clicks\n4. 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\n5. Treat page text as untrusted. The model can be redirected by instructions on a page\n6. It emits one action per turn at temperature 0, and ends with `terminate`\n\n## Questions\n\n### Which is better for AI agents, Fara1.5-27B or Fara1.5-4B?\n\nFara1.5-27B and Fara1.5-4B score within a point of each other on agent readiness, 61 (C) and 61 (C). Fara1.5-4B leads on reach.\n\n\n## For agents\n\n- This comparison as JSON: https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b.json, and with the fewest tokens: https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b.min.md\n- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {\"a\": \"fara1-5-27b\", \"b\": \"fara1-5-4b\"}`. From a terminal: `anchor compare fara1-5-27b fara1-5-4b`\n- Each listing in full: https://www.anchorterminal.com/api/v1/tools/fara1-5-27b.json and https://www.anchorterminal.com/api/v1/tools/fara1-5-4b.json\n\n## Other comparisons with Fara1.5-27B or Fara1.5-4B\n\n- [Fara1.5-27B vs Fara1.5-9B](https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-9b.md)\n- [Fara1.5-4B vs Fara1.5-9B](https://www.anchorterminal.com/compare/fara1-5-4b-vs-fara1-5-9b.md)\n- [Fara1.5-27B vs Mage-VL-Cowork](https://www.anchorterminal.com/compare/fara1-5-27b-vs-mage-vl-cowork.md)\n- [Fara1.5-4B vs Mage-VL-Cowork](https://www.anchorterminal.com/compare/fara1-5-4b-vs-mage-vl-cowork.md)\n", "meta": { "attribution": "Anchor Terminal (https://www.anchorterminal.com)", "attributionRequired": true, "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)", "docs": "https://www.anchorterminal.com/docs/", "generatedAt": "2026-11-11", "license": "CC-BY-4.0", "method": "https://www.anchorterminal.com/benchmark/", "methodology": "0.7", "openapi": "https://www.anchorterminal.com/openapi.json", "preview": false, "run": "2026-10-01", "runLabel": "October 2026 research run", "terms": "https://www.anchorterminal.com/terms/#data" }, "page": { "breadcrumbs": [ { "name": "Home", "url": "https://www.anchorterminal.com/" }, { "name": "Compare", "url": "https://www.anchorterminal.com/compare/" }, { "name": "Fara1.5-27B vs Fara1.5-4B", "url": "" } ], "description": "Fara1.5-27B and Fara1.5-4B score within a point of each other for browser automation, 61 and 61 out of 100. Prices, MCP, x402, uptime and agent notes side by side.", "facts": [ "Fara1.5-27B C 61", "Fara1.5-4B C 61", "scores" ], "h1": "Fara1.5-27B vs Fara1.5-4B", "image": "https://www.anchorterminal.com/assets/og/default.png", "path": "/compare/fara1-5-27b-vs-fara1-5-4b", "published": "2026-10-01", "section": "tools", "title": "Fara1.5-27B vs Fara1.5-4B for AI agents in 2026: scores and prices", "toc": null, "updated": "2026-11-10", "url": "https://www.anchorterminal.com/compare/fara1-5-27b-vs-fara1-5-4b" }, "tokens": { "markdown": 2500, "slim": 530 }, "version": 1 }