Benchmark Almanac

Android Bench 2.0

Claude Sonnet 5.5 40% leads since Sep 2026 Not in model cards

Coding agents complete 30 long Android tasks such as building an app from mocks, library migrations and Flutter to native conversion. Tests and an LLM visual judge grade each run.

It reports pass rate plus a continuous completion rate, and runs models in their own harnesses such as Claude Code and Codex. The task set is private to limit contamination.

Made by Google (Android Developers) Released Sep 2026 Building whole projects Tasks kept private Last model added Sep 2026 Scores refreshed 2 Nov 2026
Level
Expert Specialists with years of training, often over hours of work
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Leaderboard developer.android.com
11
Models scored on its own leaderboard

Claude Sonnet 5.5 leads with 40%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best.

  1. 40.0 Claude Sonnet 5.5
  2. 37.3 Claude Opus 5.5
  3. 29.3 GPT-6.1 Sol
  4. 28.0 GPT-6 Astra
  5. 22.7 Claude Fable 5.1
  6. 21.3 Gemini 4 Argon
  7. 12.5 Qwen 3.8 Max
  8. 12.0 Kimi K3
  9. 10.7 GPT-6 Luna
  10. 8.00 Gemini 3.8 Flash
  11. 7.30 Gemini 3.7 Flash
  1. 1 Claude Sonnet 5.5 max 40%
  2. 2 Claude Opus 5.5 max 37.3%
  3. 3 GPT-6.1 Sol max 29.3%
  4. 4 GPT-6 Astra high 28%
  5. 5 Claude Fable 5.1 high 22.7%
  6. 6 Gemini 4 Argon high 21.3%
  7. 7 Qwen 3.8 Max 12.5%
  8. 8 Kimi K3 max 12%
  9. 9 GPT-6 Luna 10.7%
  10. 10 Gemini 3.8 Flash 8.00%
  11. 11 Gemini 3.7 Flash 7.30%
#ModelScoreSettingDateSource
1 Claude Sonnet 5.5 40% max, claude-code Sep 2026 Official leaderboard
2 Claude Opus 5.5 37.3% max, claude-code Sep 2026 Official leaderboard
3 GPT-6.1 Sol 29.3% max, codex Sep 2026 Official leaderboard
4 GPT-6 Astra 28% high, codex Sep 2026 Official leaderboard
5 Claude Fable 5.1 22.7% high, claude-code Sep 2026 Official leaderboard
6 Gemini 4 Argon 21.3% high, antigravity-sdk Sep 2026 Official leaderboard
7 Qwen 3.8 Max 12.5% qwen-coder Aug 2026 Official leaderboard
8 Kimi K3 12% max, kimi-code Jul 2026 Official leaderboard
9 GPT-6 Luna 10.7% codex Sep 2026 Official leaderboard
10 Gemini 3.8 Flash 8.00% antigravity-sdk Sep 2026 Official leaderboard
11 Gemini 3.7 Flash 7.30% antigravity-sdk Aug 2026 Official leaderboard

4 models have held number one since its release in Sep 2026

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

Anthropic OpenAI
Own leaderboard 4 leaders
Claude Sonnet 5.5 40%

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% Near the ceiling (85%) Claude Sonnet 5.5 GPT-6 Astra Claude Fable 5.1 Qwen 3.8 Max 0% 20% 40% 60% 80% 100% Near the ceiling (85%) Claude Sonnet 5.5 GPT-6 Astra

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

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%
Replaces AndroidBench

More in Coding

All 243 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%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Opus 5.5 93.9%
  • 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%
  • 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