Benchmark Almanac

AndroidBench

Claude Opus 5 91.8% leads since July 2026
In use

Android development tasks from popular open-source apps: agent resolves issues or builds features, verified by test suites in emulator environments.

Android Bench 1.0 has 100 tasks drawn from real pull requests in open-source Android apps. Frontier models reached about 90%, and Google replaced it with 2.0 on September 17th, 2026. The GitHub repo is archived in favor of Harbor Hub.

Released
March 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Strongest model tested
Claude Opus 5 #10 on the capability index
Last in a launch post
August 2026
Last model added
August 2026
Scores refreshed
November 12th, 2026
32
Models scored on its own leaderboard
1
Launch post lists it, from 1 lab

Claude Opus 5 leads with 91.8% among models released by August 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. Every run is in the next chart.

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

32 models. Scroll sideways for the rest.

  1. 91.8 Claude Opus 5
  2. 90.9 Claude Fable 5
  3. 90.8 GPT-5.6 Sol
  4. 90.2 Kimi K3
  5. 87.6 GPT-5.6 Luna
  6. 87.0 Qwen 3.8 Max
  7. 86.8 GPT-5.6 Terra
  8. 80.2 GPT-5.5
  9. 76.2 Claude Sonnet 5
  10. 75.6 Gemini 3.6 Flash
  11. 74.3 Gemini 3.1 Pro
  12. 74.1 GPT-5.4
  13. 72.4 Claude Opus 4.8
  14. 72.2 GLM-5.2
  15. 71.1 Gemini 3.5 Flash
  16. 70.4 Kimi K2.7 Code
  17. 68.7 Claude Opus 4.7
  18. 67.6 Kimi K2.6
  19. 67.0 Claude Sonnet 4.6
  20. 63.6 MiniMax-M3
  21. 63.2 GLM-5.1
  22. 62.5 Gemini 3 Flash
  23. 60.8 MiMo-V2.5-Pro
  24. 59.5 DeepSeek-V4-Pro
  25. 57.7 Qwen3.7-Plus
  26. 54.7 DeepSeek-V4-Flash
  27. 54.2 Qwen3.7-Max
  28. 50.6 Gemini 3.5 Flash-Lite
  29. 45.1 Qwen3.6 27B
  30. 41.6 MiniMax-M2.7
  31. 37.1 Gemma 4 31B IT
  32. 37.0 Qwen 3.6 35B-A3B
  1. 1 Claude Opus 5 91.8%
  2. 2 Claude Fable 5 90.9%
  3. 3 GPT-5.6 Sol 90.8%
  4. 4 Kimi K3 90.2%
  5. 5 GPT-5.6 Luna 87.6%
  6. 6 Qwen 3.8 Max 87%
  7. 7 GPT-5.6 Terra 86.8%
  8. 8 GPT-5.5 80.2%
  9. 9 Claude Sonnet 5 76.2%
  10. 10 Gemini 3.6 Flash 75.6%
  11. 11 Gemini 3.1 Pro 74.3%
  12. 12 GPT-5.4 74.1%
  13. 13 Claude Opus 4.8 72.4%
  14. 14 GLM-5.2 72.2%
  15. 15 Gemini 3.5 Flash 71.1%
  16. 16 Kimi K2.7 Code 70.4%
  17. 17 Claude Opus 4.7 68.7%
  18. 18 Kimi K2.6 67.6%
  19. 19 Claude Sonnet 4.6 67%
  20. 20 MiniMax-M3 63.6%
  21. 21 GLM-5.1 63.2%
  22. 22 Gemini 3 Flash 62.5%
  23. 23 MiMo-V2.5-Pro 60.8%
  24. 24 DeepSeek-V4-Pro 59.5%
  25. 25 Qwen3.7-Plus 57.7%
  26. 26 DeepSeek-V4-Flash 54.7%
  27. 27 Qwen3.7-Max 54.2%
  28. 28 Gemini 3.5 Flash-Lite 50.6%
  29. 29 Qwen3.6 27B 45.1%
  30. 30 MiniMax-M2.7 41.6%
  31. 31 Gemma 4 31B IT 37.1%
  32. 32 Qwen 3.6 35B-A3B 37%

Source: Official AndroidBench leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 5 91.8% Jul 2026 Official leaderboard
2 Claude Fable 5 90.9% Jun 2026 Official leaderboard
3 GPT-5.6 Sol 90.8% Jul 2026 Official leaderboard
4 Kimi K3 90.2% Jul 2026 Official leaderboard
5 GPT-5.6 Luna 87.6% Jul 2026 Official leaderboard
6 Qwen 3.8 Max 87% Aug 2026 Official leaderboard
7 GPT-5.6 Terra 86.8% Jul 2026 Official leaderboard
8 GPT-5.5 80.2% Apr 2026 Official leaderboard
9 Claude Sonnet 5 76.2% Jun 2026 Official leaderboard
10 Gemini 3.6 Flash 75.6% Jul 2026 Official leaderboard
11 Gemini 3.1 Pro 74.3% Feb 2026 Official leaderboard
12 GPT-5.4 74.1% Mar 2026 Official leaderboard
13 Claude Opus 4.8 72.4% May 2026 Official leaderboard
14 GLM-5.2 72.2% Jun 2026 Official leaderboard
15 Gemini 3.5 Flash 71.1% May 2026 Official leaderboard
16 Kimi K2.7 Code 70.4% Jun 2026 Official leaderboard
17 Claude Opus 4.7 68.7% Apr 2026 Official leaderboard
18 Kimi K2.6 67.6% Apr 2026 Official leaderboard
19 Claude Sonnet 4.6 67% Feb 2026 Official leaderboard
20 MiniMax-M3 63.6% Jun 2026 Official leaderboard
21 GLM-5.1 63.2% Apr 2026 Official leaderboard
22 Gemini 3 Flash 62.5% Dec 2025 Official leaderboard
23 MiMo-V2.5-Pro 60.8% Apr 2026 Official leaderboard
24 DeepSeek-V4-Pro 59.5% Apr 2026 Official leaderboard
25 Qwen3.7-Plus 57.7% Jun 2026 Official leaderboard
26 DeepSeek-V4-Flash 54.7% Apr 2026 Official leaderboard
27 Qwen3.7-Max 54.2% May 2026 Official leaderboard
28 Gemini 3.5 Flash-Lite 50.6% Jul 2026 Official leaderboard
29 Qwen3.6 27B 45.1% Apr 2026 Official leaderboard
30 MiniMax-M2.7 41.6% Mar 2026 Official leaderboard
31 Gemma 4 31B IT 37.1% Apr 2026 Official leaderboard
32 Qwen 3.6 35B-A3B 37% Apr 2026 Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

The best score, 91.8%, costs $175 for the full run

Runs on the cost frontier score higher than any run that's cheaper. GPT-5.6 Luna comes within 4.2 points of the best for $7.20 for the full run, 24 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

DeepSeek Anthropic OpenAI Moonshot AI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $1.00 $10 $100 $1,000 Cost of the full run, log scale Claude Opus 5 GPT-5.6 Luna DeepSeek-V4-Pro DeepSeek-V4-Flash 0% 20% 40% 60% 80% 100% $1.00 $10 $100 $1,000 Cost of the full run, log scale Claude Opus 5 GPT-5.6 Luna

Source: Official AndroidBench leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 5 91.8% $175
Claude Fable 5 90.9% $144
GPT-5.6 Sol 90.8% $150
Kimi K3 90.2% $113
GPT-5.6 Luna 87.6% $7.20
Qwen 3.8 Max 87% $197
GPT-5.6 Terra 86.8% $48
GPT-5.5 80.2% $138
Claude Sonnet 5 76.2% $100
Gemini 3.6 Flash 75.6% $136
Gemini 3.1 Pro 74.3% $87
GPT-5.4 74.1% $83
Claude Opus 4.8 72.4% $88
GLM-5.2 72.2% $117
Gemini 3.5 Flash 71.1% $166
Kimi K2.7 Code 70.4% $48
Claude Opus 4.7 68.7% $97
Kimi K2.6 67.6% $49
Claude Sonnet 4.6 67% $128
MiniMax-M3 63.6% $42
GLM-5.1 63.2% $54
Gemini 3 Flash 62.5% $30
MiMo-V2.5-Pro 60.8% $9.20
DeepSeek-V4-Pro 59.5% $3.70
Qwen3.7-Plus 57.7% $19
DeepSeek-V4-Flash 54.7% $1.50
Qwen3.7-Max 54.2% $58
Gemini 3.5 Flash-Lite 50.6% $34
Qwen3.6 27B 45.1% $97
MiniMax-M2.7 41.6% $15
Gemma 4 31B IT 37.1% $10
Qwen 3.6 35B-A3B 37% $18

4 models have held number one on the official leaderboard since its release in March 2026

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

Google DeepMind OpenAI Anthropic
Own leaderboard Passed 85% in Jun 2026, by Claude Fable 5
Claude Opus 5 91.8%

Source: Official AndroidBench leaderboard

Claude Fable 5 on the official leaderboard passed 85% in June 2026

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 Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 GPT-5.5 Gemini 3.1 Pro Claude Sonnet 4.6 Gemini 3 Flash 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 GPT-5.5

Sources: Official AndroidBench leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

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.

Qwen
DateLaunch postScoreAs the post writes it
Aug 3, 2026 Qwen 3.8 Max 75.1 AndroidBench

One lab claim 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. Every benchmark.

Model Lab says Board says Gap
Qwen 3.8 Max 75.1% AndroidBench 87% Leaderboard −12 pts

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.

Google DeepMind OpenAI Anthropic
Android Bench 2.0 8 leaders
Claude Sonnet 5.5 40%
AndroidBench Passed 85% in Jun 2026, by Claude Fable 5
Claude Opus 5 91.8%
Android Bench 2.0 4 leaders
Claude Sonnet 5.5 40%

Source: Official leaderboards

Replaced by Android Bench 2.0

More in Coding

All 282 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471