Benchmark Almanac

ScreenSpot-v2

Seed1.5-VL 95.2% leads since May 2025 Fading

Models locate a described element on mobile, desktop and web screenshots. A click inside the target box counts as correct.

Released with OS-Atlas as a relabeled ScreenSpot that fixes wrong annotations, so scores differ from the original.

Made by OS-Atlas team (Shanghai AI Lab and others) Released Oct 2024 Finding things on screen All tasks public Last in a launch post Aug 2026 , dropped after it saturated Last model added Oct 2026
Level
Everyday Most adults, without training
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Short answer A number, word or short string matched to a key.
Paper
14
Models scored in lab cards
9
Launch posts list it, from 7 labs
12
System cards report it, from 8 labs

Seed1.5-VL leads with 95.2% among models released by Aug 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

The newest model with a score here came out in Aug 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

14 models. Scroll sideways for the rest.

  1. 95.2 Seed1.5-VL
  2. 94.4 Fara1.5-27B
  3. 94.2 UI-TARS-1.5
  4. 93.6 Intern-S1-Pro
  5. 93.3 Muse Glimmer
  6. 92.9 InternVL3.5-241B-A28B
  7. 92.8 Nemotron 3 Nano Omni 30B-A3B
  8. 92.6 STEP3-VL-10B
  9. 91.4 Kimi-VL-A3B-Thinking-2506
  10. 91.2 LongCat-Flash-Omni
  11. 89.3 Fara-7B
  12. 88.2 Phi-4-Reasoning-Vision-15B
  13. 80.7 LFM2.5-VL-3B
  14. 42.8 Nemotron Nano 12B v2 VL
  1. 1 Seed1.5-VL 95.2%
  2. 2 Fara1.5-27B 94.4%
  3. 3 UI-TARS-1.5 94.2%
  4. 4 Intern-S1-Pro 93.6%
  5. 5 Muse Glimmer 93.3%
  6. 6 InternVL3.5-241B-A28B 92.9%
  7. 7 Nemotron 3 Nano Omni 30B-A3B 92.8%
  8. 8 STEP3-VL-10B 92.6%
  9. 9 Kimi-VL-A3B-Thinking-2506 91.4%
  10. 10 LongCat-Flash-Omni 91.2%
  11. 11 Fara-7B 89.3%
  12. 12 Phi-4-Reasoning-Vision-15B 88.2%
  13. 13 LFM2.5-VL-3B 80.7%
  14. 14 Nemotron Nano 12B v2 VL 42.8%
#ModelScoreSettingDateSource
1 Seed1.5-VL 95.2% May 2025 Model card
2 Fara1.5-27B 94.4% Jun 2026 Model card
3 UI-TARS-1.5 94.2% Apr 2025 Model card
4 Intern-S1-Pro 93.6% Feb 2026 Model card
5 Muse Glimmer 93.3% Oct 2026 Model card
6 InternVL3.5-241B-A28B 92.9% Aug 2025 Model card
7 Nemotron 3 Nano Omni 30B-A3B 92.8% Apr 2026 Model card
8 STEP3-VL-10B 92.6% Jan 2026 Model card
9 Kimi-VL-A3B-Thinking-2506 91.4% Jun 2025 Model card
10 LongCat-Flash-Omni 91.2% Oct 2025 Model card
11 Fara-7B 89.3% Nov 2025 Model card
12 Phi-4-Reasoning-Vision-15B 88.2% Mar 2026 Model card
13 LFM2.5-VL-3B 80.7% Aug 2026 Model card
14 Nemotron Nano 12B v2 VL 42.8% Nov 2025 Model card

2 models have held number one since its release in Oct 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

ByteDance Seed
Lab claims Passed 85% in Apr 2025, by UI-TARS-1.5
Seed1.5-VL 95.2%

The best score passed 85% in Apr 2025

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Seed1.5-VL 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Seed1.5-VL

Labs headline it less often now, last in Aug 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.

ByteDance Seed
Moonshot AI
Shanghai AI Lab
Meituan
StepFun
Microsoft
Liquid AI
DateLaunch postScoreAs the post writes it
12 Aug 2026 LFM2.5-VL-3B 80.7 Mean of ScreenSpot-v2 Desktop 78.7 / Mobile 81.2 / Web 82.2
4 Mar 2026 Phi-4-Reasoning-Vision-15B 88.2 ScreenSpot_v2 (default mode)
2 Feb 2026 Intern-S1-Pro 93.6 ScreenSpot V2
13 Jan 2026 STEP3-VL-10B 92.6 ScreenSpot-V2
31 Oct 2025 LongCat-Flash-Omni 91.2 ScreenSpot-v2
25 Aug 2025 InternVL3.5-241B-A28B 92.9 ScreenSpot-v2
21 Jun 2025 Kimi-VL-A3B-Thinking-2506 91.4 ScreenSpot-V2 (Acc)
12 May 2025 Seed1.5-VL 95.2 ScreenSpot-V2
16 Apr 2025 UI-TARS-1.5 94.2 ScreensSpot-V2

12 system cards report it, from 8 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.

DateDocumentScoreAs the document writes itSection
19 Oct 2026 Muse Glimmer technical report 93.3 Screenspot v2, high (Table 5) Capabilities
18 Jun 2026 Fara1.5-27B technical report 94.4 ScreenSpot-v2 grounding accuracy (Table 6) Capabilities
27 Apr 2026 Nemotron 3 Nano Omni 30B-A3B technical report 92.8 ScreenSpot-v2, reasoning on, 93.4 reasoning off (Table 7) Capabilities
26 Mar 2026 Intern-S1-Pro technical report 93.6 ScreenSpot V2, non-thinking (Table 2) Capabilities
4 Mar 2026 Phi-4-Reasoning-Vision-15B technical report 88.2 ScreenSpotv2, default mode, 88.3 force nothink, 88.1 force thinking (Tables 5 and 6) Capabilities
14 Jan 2026 STEP3-VL-10B technical report 92.6 ScreenSpot-V2, SeRe (Table 1) Capabilities
24 Nov 2025 Fara-7B technical report 89.3 ScreenSpot-V2 (Table 13a) Capabilities
6 Nov 2025 Nemotron Nano 12B v2 VL technical report 42.8 ScreenSpot-v2, reasoning on (Table 3). Reasoning off 41.7 Capabilities
31 Oct 2025 LongCat-Flash-Omni technical report 91.2 ScreenSpot-v2 (Table 5) Capabilities
25 Aug 2025 InternVL3.5-241B-A28B technical report 92.9 ScreenSpot-v2 (Table 2, Table 10) Capabilities
23 Jun 2025 Kimi-VL-A3B-Thinking-2506 technical report 91.4 ScreenSpot-V2 (Acc) (Table 5) Capabilities
11 May 2025 Seed1.5-VL technical report 95.2 ScreenSpot-V2, thinking (Table 8) Capabilities

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

Qwen Amazon ByteDance Seed
ScreenSpot-v2 Passed 85% in Apr 2025, by UI-TARS-1.5
Seed1.5-VL 95.2%
ScreenSpot Passed 85% in Jan 2025, by Qwen2.5-VL-72B
Qwen3-VL-235B-A22B 95.4%
ScreenSpot-v2 Passed 85% in Apr 2025, by UI-TARS-1.5
Seed1.5-VL 95.2%
Replaces ScreenSpot

More in Computer and browser use

All 56 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. Kimi K3 74.4%