BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases8

FrontierCode 2.0 (AA)

As of 28 Oct 2026, the top score on FrontierCode 2.0 (AA) is 46.4% by Claude Opus 5.5 (Anthropic), ahead of Claude Opus 5.5 at 43.8%, across 77 model configurations with a published result (Artificial Analysis, result dated 22 Sept 2026). FrontierCode 2.0: 132 real open-source repository coding tasks from Cognition, graded on whether a maintainer would merge the change. Used in Artificial Analysis' Intelligence Index v5. FrontierCode 2.0 (AA) counts toward the BenchLeader coding score.

Key facts (as of 28 Oct 2026)

  • Claude Opus 5.5 holds the top FrontierCode 2.0 (AA) score at 46.4% (Artificial Analysis, 22 Sept 2026).
  • Claude Opus 5.5 is second on FrontierCode 2.0 (AA) at 43.8% (Artificial Analysis, 22 Sept 2026).
  • FrontierCode 2.0 (AA): Cognition's FrontierCode 2.0: 132 real open-source repository tasks graded on whether a maintainer would merge the change. Used in Intelligence Index v5. Run by Artificial Analysis (Artificial Analysis).
  • 77 model configurations have a published FrontierCode 2.0 (AA) result on BenchLeader (BenchLeader, 28 Oct 2026).
  • FrontierCode 2.0 (AA) feeds the coding category of the BenchLeader Index (BenchLeader methodology).

Cognition's FrontierCode 2.0: 132 real open-source repository tasks graded on whether a maintainer would merge the change. Used in Intelligence Index v5. Run by Artificial Analysis.

Published by
Artificial Analysis
Category
Coding
Index weight
1.0
Models
77
Data as of
28 Oct 2026
Newest result
26 Oct 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

What the test looks like

FrontierCode 2.0: 132 real open-source repository coding tasks from Cognition, graded on whether a maintainer would merge the change. Used in Artificial Analysis' Intelligence Index v5.

How it is scored

Accuracy percentage. Run by Artificial Analysis.

What to keep in mind

Distinct from FrontierCode 1.1 and from Epoch's FrontierCode run; far fewer models have been run on 2.0 so far.

Top 15
  1. 1Claude Opus 5.5 (max)46.4%
  2. 2Claude Opus 5.5 (xhigh)43.8%
  3. 3GPT-6 Astra (max)42.4%
  4. 4GPT-6.1 Terra (max)42.3%
  5. 5Claude Sonnet 5.5 (max)41.7%
  6. 6GPT-6 Astra (xhigh)41.4%
  7. 7GPT-6.1 Sol (max)40.6%
  8. 8GPT-6 Astra (high)40.1%
  9. 9GPT-6.1 Sol (xhigh)38.5%
  10. 10Muse Wonder Full (max)37.8%
  11. 11GPT-6.1 Terra (high)37.5%
  12. 12GPT-6.1 Terra (xhigh)36.9%
  13. 13GPT-6.1 Sol (high)36.7%
  14. 14GPT-6 Astra (medium)36.6%
  15. 15Claude Opus 5.5 (high)35.9%
77 of 77
#
1Claude Opus 5.5max46.4%69.82026-09-22
2Claude Opus 5.5xhigh43.8%66.82026-09-22
3GPT-6 Astramax42.4%68.32026-09-03
4NewGPT-6.1 Terramax42.3%65.02026-10-26
5Claude Sonnet 5.5max41.7%65.72026-09-28
6GPT-6 Astraxhigh41.4%67.12026-09-03
7GPT-6.1 Solmax40.6%67.62026-09-29
8GPT-6 Astrahigh40.1%67.32026-09-03
9GPT-6.1 Solxhigh38.5%64.82026-09-29
10Muse Wonder Fullmax37.8%62.72026-10-12
11NewGPT-6.1 Terrahigh37.5%59.92026-10-26
12NewGPT-6.1 Terraxhigh36.9%60.42026-10-26
13GPT-6.1 Solhigh36.7%64.32026-09-29
14GPT-6 Astramedium36.6%65.22026-09-03
15Claude Opus 5.5high35.9%65.92026-09-22
16Claude Fable 5.1xhigh35.7%65.02026-09-01
17Claude Sonnet 5.5xhigh35.6%63.42026-09-28
18Claude Haiku 5.5max34.9%56.42026-10-07
19NewGPT-6.1 Terramedium34.5%58.32026-10-26
20GPT-6.1 Solmedium33.4%63.12026-09-29
21NewGPT-6.1 Terrano reasoning33.4%51.62026-10-26
22Claude Opus 5.5medium33.1%62.22026-09-22
23Claude Fable 5.1max31.6%66.92026-09-01
24Claude Fable 5.1high30.3%63.52026-09-01
25Claude Haiku 5.5xhigh29.9%56.02026-10-07
26Claude Fable 5.1medium28.8%61.52026-09-01
27Grok 4.7xhigh28.7%57.42026-09-21
28DeepSeek V4.1 Flash 1017max28.5%57.82026-10-17
29Muse Wondermax27.9%61.22026-10-12
30Grok 4.7high27.8%56.12026-09-21
31GPT-6 Astralow26.2%62.02026-09-03
32Claude Sonnet 5.5high25.7%57.92026-09-28
33Gemini 4 Argonhigh25.7%66.02026-09-30
34NewKimi K3.1 Pro25.4%59.82026-10-23
35GPT-6 Lunamax25.2%55.42026-09-22
36NewGPT-6.1 Terralow24.5%54.92026-10-26
37MiMo-V2.6-Pro24.0%56.92026-09-21
38Qwen3.8 Maxmax23.4%59.12026-09-02
39Claude Haiku 5.5high23.2%54.52026-10-07
40Muse Spark 1.3max23.1%60.92026-09-02
41Claude Fable 5.1low22.8%58.12026-09-01
42MiniMax M321.9%54.32026-06-01
43Grok 4.7low21.8%54.42026-09-21
44MiMo-V2.6-Flash21.6%53.82026-09-21
45GLM 5.3max21.6%57.62026-08-18
46MiniMax M3.1 Flash20.8%52.62026-10-19
47GPT-6.1 Sollow20.2%59.02026-09-29
48Gemini 3.8 Flashhigh18.7%60.32026-09-02
49GPT-6 Lunahigh18.6%49.22026-09-22
50GPT-6 Lunaxhigh18.4%50.22026-09-22
51DeepSeek V4 Promax17.1%54.52026-08-13
52Ling 3.1 Flash17.0%49.82026-10-22
53Gemini 3.8 Flashmedium16.5%58.62026-09-02
54Claude Haiku 5.5medium16.1%48.52026-10-07
55Claude Opus 5.5low15.8%57.82026-09-22
56Qwen3.8 27Bxhigh14.9%50.22026-08-14
57Mistral Large 413.6%51.42026-10-06
58GPT-6 Lunamedium12.8%47.72026-09-22
59GLM 5.3low12.7%45.92026-08-18
60Step 5 Preview12.0%55.02026-09-18

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “FrontierCode 2.0 (AA) leaderboard”, data as of 28 Oct 2026, https://www.benchleader.com/benchmarks/aa_frontiercode_20.

FrontierCode 2.0 (AA): questions

What does FrontierCode 2.0 (AA) measure?
FrontierCode 2.0: 132 real open-source repository coding tasks from Cognition, graded on whether a maintainer would merge the change. Used in Artificial Analysis' Intelligence Index v5. Scores are reported in percent of tasks solved; higher is better.
What is the top score on FrontierCode 2.0 (AA)?
As of 28 Oct 2026, the highest score on FrontierCode 2.0 (AA) is 46.4% by Claude Opus 5.5 (result dated 22 Sept 2026), ahead of Claude Opus 5.5 at 43.8%. Scores come from Artificial Analysis and are re-read daily.
Which AI model leads FrontierCode 2.0 (AA)?
Claude Opus 5.5 leads FrontierCode 2.0 (AA) with 46.4% as of 28 Oct 2026, ahead of Claude Opus 5.5 at 43.8% (result dated 22 Sept 2026).
How many models have FrontierCode 2.0 (AA) results?
77 model configurations have a FrontierCode 2.0 (AA) result on BenchLeader, all taken from Artificial Analysis.
Who runs FrontierCode 2.0 (AA) and how often is it updated?
FrontierCode 2.0 (AA) is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
Does FrontierCode 2.0 (AA) count toward the BenchLeader Index?
Yes. FrontierCode 2.0 (AA) contributes to the coding category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.