Benchmark Almanac

BrowseComp Extended

Claude Sonnet 5.5 73% leads since September 2026
Not in model cards

Agents browse the web to find hard-to-locate facts with short answers. Mercor adds 100 new expert-written tasks to the 130 public ones and reports both scores.

Held-out extension of browsecomp. Mercor shows the original and extended score for each model. The month is approximate. The Wayback Machine first saved the earliest extended pages on August 15th, 2026.

Made by
Mercor
Released
August 2026
Kind
Web browsing
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
Claude Opus 5.5 #4 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
43
Entries on its own leaderboard, 42 models we track

Claude Sonnet 5.5 leads with 73%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

43 entries. Scroll sideways for the rest.

  1. 73.0 Claude Sonnet 5.5
  2. 72.0 Gemini 4 Argon
  3. 67.7 GPT-6.1 Sol
  4. 63.3 Claude Fable 5.1
  5. 63.2 GPT-6 Sol
  6. 60.5 Claude Opus 5.5
  7. 48.2 DeepSeek V4.1 Flash
  8. 47.5 Muse Wonder
  9. 47.2 GPT-6 Luna
  10. 44.5 Claude Fable 5
  11. 44.0 Claude Opus 5
  12. 41.5 Gemini 3.8 Flash
  13. 39.5 Grok 4.7
  14. 32.5 Muse Spark 1.2
  15. 32.3 GPT-5.6 Sol
  16. 32.0 Muse Spark Open
  17. 30.0 Claude Opus 4.8
  18. 28.5 Kimi K3
  19. 27.5 Claude Haiku 5.5
  20. 27.0 DeepSeek V4 Pro 0813
  21. 25.5 Gemini 3.7 Flash
  22. 25.3 GPT-5.6 Terra
  23. 25.0 Beam
  24. 23.0 GLM-5.2
  25. 22.0 DeepSeek-V4-Flash
  26. 21.5 Qwen 3.8 Max
  27. 21.5 Grok 4.6
  28. 19.0 GPT-5.5
  29. 19.0 Claude Sonnet 5
  30. 16.5 GPT-5.6 Luna
  31. 14.2 Claude Sonnet 4.6
  32. 14.2 Claude Opus 4.7
  33. 14.2 Grok 4.5
  34. 14.0 GPT-5.4
  35. 13.8 Gemini 3.6 Flash
  36. 10.7 Gemini 3.1 Pro
  37. 8.30 DeepSeek-V4-Pro
  38. 8.00 Kimi K2.7 Code
  39. 7.70 MiniMax-M3
  40. 4.50 Inkling
  41. 3.30 DeepSeek-V3.2
  42. 0.70 Gemini 3.5 Flash
  43. 0.50 MiniMax-M2.7
  1. 1 Claude Sonnet 5.5 max 73%
  2. 2 Gemini 4 Argon high 72%
  3. 3 GPT-6.1 Sol max 67.7%
  4. 4 Claude Fable 5.1 max 63.3%
  5. 5 GPT-6 Sol max 63.2%
  6. 6 Claude Opus 5.5 max 60.5%
  7. 7 DeepSeek V4.1 Flash max 48.2%
  8. 8 Muse Wonder max 47.5%
  9. 9 GPT-6 Luna max 47.2%
  10. 10 Claude Fable 5 high 44.5%
  11. 11 Claude Opus 5 max 44%
  12. 12 Gemini 3.8 Flash high 41.5%
  13. 13 Grok 4.7 xhigh 39.5%
  14. 14 Muse Spark 1.2 xhigh 32.5%
  15. 15 GPT-5.6 Sol xhigh 32.3%
  16. 16 Muse Spark Open xhigh 32%
  17. 17 Claude Opus 4.8 max 30%
  18. 18 Kimi K3 max 28.5%
  19. 19 Claude Haiku 5.5 max 27.5%
  20. 20 DeepSeek V4 Pro 0813 max 27%
  21. 21 Gemini 3.7 Flash high 25.5%
  22. 22 GPT-5.6 Terra max 25.3%
  23. 23 Beam max 25%
  24. 24 GLM-5.2 max 23%
  25. 25 DeepSeek-V4-Flash max 22%
  26. 26 Qwen 3.8 Max xhigh 21.5%
  27. 27 Grok 4.6 xhigh 21.5%
  28. 28 GPT-5.5 xhigh 19%
  29. 29 Claude Sonnet 5 max 19%
  30. 30 GPT-5.6 Luna max 16.5%
  31. 31 Claude Sonnet 4.6 high 14.2%
  32. 32 Claude Opus 4.7 max 14.2%
  33. 33 Grok 4.5 high 14.2%
  34. 34 GPT-5.4 xhigh 14%
  35. 35 Gemini 3.6 Flash high 13.8%
  36. 36 Gemini 3.1 Pro high 10.7%
  37. 37 DeepSeek-V4-Pro max 8.30%
  38. 38 Kimi K2.7 Code high 8.00%
  39. 39 MiniMax-M3 high 7.70%
  40. 40 Inkling high 4.50%
  41. 41 DeepSeek-V3.2 3.30%
  42. 42 Gemini 3.5 Flash high 0.70%
  43. 43 MiniMax-M2.7 high 0.50%

Source: Official BrowseComp Extended leaderboard

#EntryScoreSettingDateSource
1 Claude Sonnet 5.5 73% max Sep 2026 Official leaderboard
2 Gemini 4 Argon 72% high Sep 2026 Official leaderboard
3 GPT-6.1 Sol 67.7% max Sep 2026 Official leaderboard
4 Claude Fable 5.1 63.3% max Sep 2026 Official leaderboard
5 GPT-6 Sol 63.2% max Sep 2026 Official leaderboard
6 Claude Opus 5.5 60.5% max Sep 2026 Official leaderboard
7 DeepSeek V4.1 Flash 48.2% max Sep 2026 Official leaderboard
8 Muse Wonder 47.5% max Oct 2026 Official leaderboard
9 GPT-6 Luna 47.2% max Sep 2026 Official leaderboard
10 Claude Fable 5 44.5% high Jun 2026 Official leaderboard
11 Claude Opus 5 44% max Jul 2026 Official leaderboard
12 Gemini 3.8 Flash 41.5% high Sep 2026 Official leaderboard
13 Grok 4.7 39.5% xhigh Sep 2026 Official leaderboard
14 Muse Spark 1.2 32.5% xhigh Aug 2026 Official leaderboard
15 GPT-5.6 Sol 32.3% xhigh Jul 2026 Official leaderboard
16 Muse Spark Open 32% xhigh Oct 2026 Official leaderboard
17 Claude Opus 4.8 30% max May 2026 Official leaderboard
18 Kimi K3 28.5% max Jul 2026 Official leaderboard
19 Claude Haiku 5.5 27.5% max Oct 2026 Official leaderboard
20 DeepSeek V4 Pro 0813 27% max Aug 2026 Official leaderboard
21 Gemini 3.7 Flash 25.5% high Aug 2026 Official leaderboard
22 GPT-5.6 Terra 25.3% max Jul 2026 Official leaderboard
23 Beam 25% max Official leaderboard
24 GLM-5.2 23% max Jun 2026 Official leaderboard
25 DeepSeek-V4-Flash 22% max Apr 2026 Official leaderboard
26 Qwen 3.8 Max 21.5% xhigh Aug 2026 Official leaderboard
27 Grok 4.6 21.5% xhigh Aug 2026 Official leaderboard
28 GPT-5.5 19% xhigh Apr 2026 Official leaderboard
29 Claude Sonnet 5 19% max Jun 2026 Official leaderboard
30 GPT-5.6 Luna 16.5% max Jul 2026 Official leaderboard
31 Claude Sonnet 4.6 14.2% high Feb 2026 Official leaderboard
32 Claude Opus 4.7 14.2% max Apr 2026 Official leaderboard
33 Grok 4.5 14.2% high Jul 2026 Official leaderboard
34 GPT-5.4 14% xhigh Mar 2026 Official leaderboard
35 Gemini 3.6 Flash 13.8% high Jul 2026 Official leaderboard
36 Gemini 3.1 Pro 10.7% high Feb 2026 Official leaderboard
37 DeepSeek-V4-Pro 8.30% max Apr 2026 Official leaderboard
38 Kimi K2.7 Code 8.00% high Jun 2026 Official leaderboard
39 MiniMax-M3 7.70% high Jun 2026 Official leaderboard
40 Inkling 4.50% high Jul 2026 Official leaderboard
41 DeepSeek-V3.2 3.30% Dec 2025 Official leaderboard
42 Gemini 3.5 Flash 0.70% high May 2026 Official leaderboard
43 MiniMax-M2.7 0.50% high Mar 2026 Official leaderboard

3 models have held number one on the official leaderboard since its release in August 2026

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

Anthropic
Own leaderboard 3 leaders
Claude Sonnet 5.5 73%

Source: Official BrowseComp Extended leaderboard

Scores over time

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

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Sonnet 5.5 Claude Fable 5.1 Claude Fable 5 Claude Opus 4.8 DeepSeek-V4-Flash Claude Sonnet 4.6 DeepSeek-V3.2 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Sonnet 5.5 Claude Fable 5.1

Source: Official BrowseComp Extended leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Computer and browser use

All 69 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%