WebTailBench
Computer-use agents complete 609 hand-verified tasks on live websites across 11 categories, such as bookings, real estate and job search. A companion set of 111 harmful tasks tests refusals.
Released with Fara-7B. On May 12th, 2026 Microsoft added a V2 split that refreshes calendar-bound task strings and rubrics for the same 609 tasks. The maker does not say V2 scores differ, so it stays one record.
- Made by
- Microsoft
- Released
- November 2025
- Kind
- Web browsing
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Strongest model tested
- GPT-5.4 #27 on the capability index
- Last in a launch post
- May 2026, dropped before it saturated
- Last model added
- July 2026
- Scores refreshed
- November 12th, 2026
GPT-5.4 leads with 57.4% among models released by July 2026
Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.
GPT-5.4 leads with 57.4%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 GPT-5.4 v1.5 57.4%
- 2 GPT-5 v1.5 45.1%
- 3
Fara1.5-27B v1.5 40.2% - 4 o3 v1.5 35%
- 5 Fara1.5-9B v1.5 32.3%
- 6 Fara1.5-4B v1.5 27.4%
- 7 Fara-7B v1.5 24.1%
Source: Official WebTailBench leaderboard
- 1 GPT-5.4 v1.5 57.4%
- 2
MolmoWeb 8B 49.5% - 3 GPT-5 v1.5 45.1%
- 4 Fara1.5-27B v1.5 40.2%
- 5 Fara-7B 38.4%
- 6 o3 v1.5 35%
- 7 Fara1.5-9B v1.5 32.3%
- 8 Fara1.5-4B v1.5 27.4%
Sources: Official WebTailBench leaderboard and labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.4 | 57.4% | v1.5, FaraGen1.5 Solver | Mar 2026 | Official leaderboard |
| 2 | GPT-5 | 45.1% | v1.5, SoM | Aug 2025 | Official leaderboard |
| 3 | Fara1.5-27B | 40.2% | v1.5 | May 2026 | Official leaderboard |
| 4 | o3 | 35% | v1.5, SoM | Apr 2025 | Official leaderboard |
| 5 | Fara1.5-9B | 32.3% | v1.5 | Jul 2026 | Official leaderboard |
| 6 | Fara1.5-4B | 27.4% | v1.5 | Jul 2026 | Official leaderboard |
| 7 | Fara-7B | 24.1% | v1.5 | Nov 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.4 | 57.4% | v1.5, FaraGen1.5 Solver | Mar 2026 | Official leaderboard |
| 2 | MolmoWeb 8B | 49.5% | Mar 2026 | Lab model card | |
| 3 | GPT-5 | 45.1% | v1.5, SoM | Aug 2025 | Official leaderboard |
| 4 | Fara1.5-27B | 40.2% | v1.5 | May 2026 | Official leaderboard |
| 5 | Fara-7B | 38.4% | Nov 2025 | Lab model card | |
| 6 | o3 | 35% | v1.5, SoM | Apr 2025 | Official leaderboard |
| 7 | Fara1.5-9B | 32.3% | v1.5 | Jul 2026 | Official leaderboard |
| 8 | Fara1.5-4B | 27.4% | v1.5 | Jul 2026 | Official leaderboard |
Labs report scores for 3 models in their cards. Switch the scores above to rank them together.
2 models have held number one on the official leaderboard since its release in November 2025 2 models have held number one with claims included since its release in November 2025
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official WebTailBench leaderboard and labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official WebTailBench leaderboard and labs’ model cards
Labs headline it less often now, last in May 2026
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| May 21, 2026 | Fara1.5-27B | 40.2 | WebTailBench |
| Mar 24, 2026 | MolmoWeb 8B | 49.5 | WebTailBench |
| Nov 24, 2025 | Fara-7B | 38.4 | WebTailBench |
3 system cards report it, from 2 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Jun 18, 2026 | Fara1.5-27B technical report | 40.2 | WebTailBench v1.5 Outcome Success, refreshed tasks and adapted critical-point verifier, 95% CI 3.0. Process Success 72.9 (Tables 3 and 7) | Capabilities |
| Apr 9, 2026 | MolmoWeb 8B technical report | 49.5 | WebTailBench, scored with the WebVoyager judge, 100 steps (Table 4) | Capabilities |
| Nov 24, 2025 | Fara-7B technical report | 38.4 | WebTailBench success rate, mean of 3 runs, macro and micro average both 38.4 over 11 segments (Table 9, Table 11) | Capabilities |
2 lab claims next to the leaderboard
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.
More in Computer and browser use
- BrowseComp In use Hard-to-find facts that need persistent web browsing. Short verifiable answers graded for exact correctness. Claude Fable 5 93.3%
- ScreenSpot-Pro In use GUI grounding on high-resolution professional-software screenshots. The model must locate the target UI element. GPT-6 Astra 92.7%
- AndroidWorld In use Agents complete 116 tasks across 20 real Android apps in an emulator. Each task checks the device's system state to judge success. GPT-6 Astra 100%
- OSWorld 2.1 In use Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. Claude Opus 5.5 81.8%
- WebArena-Verified In use Agents complete 812 audited WebArena tasks on self-hosted websites. Deterministic evaluators check the agent's answer and captured network traces.
Muse Spark 1.1 69% - MobileWorld In use Agents complete 201 tasks across 20 mobile apps, including tasks that need user clarification or MCP tools. Backend checks verify success. GPT-5 55.3%