Frontier Log

Benchmark · Active

FrontierCode 2.0

Created by Cognition, 27 Oct 2026 · Higher is better · Website

Versions

VersionReleasedStatusModels
FrontierCode 1.17 Jul 2026Active15
FrontierCode 2.027 Oct 2026Active31

Settings

Task setScoringReported by
All tasks (132)ScoreOpenAIAnthropicGoogleMetaCognition’s leaderboard
All tasks (132)Pass rateCognition’s leaderboard

Bold: Benchmark default based on Leaderboard and Announcement · Download image

Results

Show
0102030405060Apr 2026Jul 2026Oct 2026Jan 202749.1%Fable 5.227.2%Muse Ember45.5%Opus 5.532.4%Fable 5.128.9%GPT-5.6 Sol23.6%GPT-5.51.3%SWE-1.60102030405060Apr 2026Jul 2026Oct 2026Jan 202749.1%Fable 5.227.2%Muse Ember45.5%Opus 5.532.4%Fable 5.128.9%GPT-5.6 Sol
  • OpenAI
  • Anthropic
  • Google
  • Other labs
  • Reported in a later post or card, or on the creator’s leaderboard
ModelHarness and effortScoreSource
Claude Fable 5.29 Nov 2026Claude Codexhigh effort49.1%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 2 other documents: 1 2Run by Cognition · mean@5, per the system cardScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codelow effort34.2%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 2 other documents: 1 2Run by Cognition · mean@5, per the system card · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codemedium effort37.6%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 2 other documents: 1 2Run by Cognition · mean@5, per the system card · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codehigh effort45.3%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 2 other documents: 1 2Run by Cognition · mean@5, per the system card · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codemax effort45.6%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codemax effort45.7%Claude Fable 5.2 and Claude Mythos 5.2 System CardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · mean@5 · includes the graded scores of 22 attempts (out of 660) that timed out · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.29 Nov 2026Claude Codexhigh effort50.2%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
Gemini 4 Neon18 Nov 2026Harness not specifiedhigh thinking46.2%Introducing Gemini 4 NeonCorroborated The same number is printed in 4 other documents. The first 3 are linked: 1 2 3Run by Cognition · set and measure of the official public leaderboard the evaluation methodology citesSource image
Gemini 4 Neon18 Nov 2026Antigravity CLIhigh effort46.2%FrontierCode LeaderboardCorroborated The same number is printed in 1 other document: 1Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
Claude Opus 5.522 Sep 2026Claude Codehighest-available effort, max effort in the post's FrontierCode 2.0 chart45.5%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 2 other documents: 1 2Run by Cognition · mean@5, per the system card · reported laterScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Opus 5.522 Sep 2026Claude Codelow effort14.7%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Opus 5.522 Sep 2026Claude Codemedium effort30.1%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Opus 5.522 Sep 2026Claude Codehigh effort34.5%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Opus 5.522 Sep 2026Claude Codexhigh effort43.4%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Opus 5.522 Sep 2026Claude Codemax effort47.3%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
GPT-6.2 Astra17 Nov 2026Harness not specifiedmax effort44.2%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1Run by CognitionOur capture
GPT-6.2 Astra17 Nov 2026Harness not specifiedlow effort38.1%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Astra17 Nov 2026Harness not specifiedmedium effort42.2%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Astra17 Nov 2026Harness not specifiedhigh effort43.6%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Astra17 Nov 2026Harness not specifiedxhigh effort44.1%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Astra17 Nov 2026Codexmax effort44.2%FrontierCode LeaderboardCorroborated The same number is printed in 1 other document: 1Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
GPT-6.2 Sol17 Nov 2026Harness not specifiedmax effort43.8%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1Run by CognitionOur capture
GPT-6.2 Sol17 Nov 2026Harness not specifiedlow effort33.1%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Sol17 Nov 2026Harness not specifiedmedium effort38.3%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Sol17 Nov 2026Harness not specifiedhigh effort39.7%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Sol17 Nov 2026Harness not specifiedxhigh effort39.8%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · another reportOur capture
GPT-6.2 Sol17 Nov 2026Codexmax effort43.8%FrontierCode LeaderboardCorroborated The same number is printed in 1 other document: 1Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
GPT-6 Astra3 Sep 2026Harness not specifiedmax effort42.8%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 2 other documents: 1 2reported laterScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6 Astra3 Sep 2026Harness not specifiedlow effort31.1%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1Run by Cognition · reported later · another reportOur capture
GPT-6 Astra3 Sep 2026Harness not specifiedmedium effort36.8%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6 Astra3 Sep 2026Harness not specifiedhigh effort38.3%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6 Astra3 Sep 2026Harness not specifiedxhigh effort41.4%Introducing GPT-6.2 Astra and SolCorroborated The same number is printed in 1 other document: 1reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6 Astra3 Sep 2026Codexmax effort44.0%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
Gemini 4 Carbon28 Oct 2026Antigravity CLIhigh effort42.6%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Claude Sonnet 5.528 Sep 2026Claude Codemax effort42.2%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Harness not specifiedmax effort39.0%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.reported laterScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Harness not specifiedlow effort25.6%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Harness not specifiedmedium effort29.6%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Harness not specifiedhigh effort34.8%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Harness not specifiedxhigh effort32.3%Introducing GPT-6.2 Astra and SolChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026Our capture
GPT-6.1 Sol29 Sep 2026Codexmax effort39.6%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
Gemini 4 Hydrogen12 Nov 2026Antigravity CLIxhigh effort38.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GPT-6 Sol22 Sep 2026Codexmax effort34.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Claude Fable 5.11 Sep 2026Claude Codexhigh effort32.4%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported laterScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.11 Sep 2026Claude Codelow effort20.6%Introducing Claude Fable 5.2 and Claude Mythos 5.2Checked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.11 Sep 2026Claude Codemedium effort21.5%Introducing Claude Fable 5.2 and Claude Mythos 5.2Checked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.11 Sep 2026Claude Codehigh effort25.7%Introducing Claude Fable 5.2 and Claude Mythos 5.2Checked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.11 Sep 2026Claude Codemax effort31.7%Introducing Claude Fable 5.2 and Claude Mythos 5.2Corroborated The same number is printed in 1 other document: 1Run by Cognition · mean@5, per the system card · reported later · another reportScored before Cognition updated its cheating scanner to remove false positives and rescored all results on 11 Nov 2026
Claude Fable 5.11 Sep 2026Claude Codexhigh effort34.0%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026 · another reportOur capture
Claude Haiku 5.57 Oct 2026Claude Codemax effort28.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GPT-5.6 Sol9 Jul 2026Codexmax effort28.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Claude Opus 524 Jul 2026Claude Codehigh effort28.1%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Gemini 4 Argon30 Sep 2026Devin CLIhigh effort27.6%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Muse Ember9 Nov 2026Stirrupmax reasoning effort27.2%Introducing Our Smallest, Most Efficient Model: Muse EmberChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Run by Artificial Analysis · 3 attempts per task, averaged, per Artificial Analysis’s methodology · 500-turn cap, per Artificial Analysis’s methodology · GPT-6 Luna (medium) LLM judge, per Artificial Analysis’s methodology · Setting from Artificial Analysis’s methodologyOur capture
SWE-210 Sep 2026Devin CLIhigh effort24.1%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
SWE-2 Flash4 Nov 2026Devin CLImax effort24.0%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GPT-5.523 Apr 2026Codexxhigh effort23.6%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GPT-6 Luna22 Sep 2026Codexmax effort22.7%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Inkling 2 Small5 Nov 2026Devin CLImax effort22.6%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
DeepSeek-V4.1-Flash-101717 Oct 2026Devin CLImax effort20.2%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GLM-5.314 Aug 2026Devin CLImax effort20.2%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Kimi K3.1 Pro24 Oct 2026Devin CLImax effort19.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Mistral Large 46 Oct 2026Mistral Vibehigh effort19.7%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Kimi K316 Jul 2026Devin CLIhigh effort18.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
SWE-1.78 Jul 2026Devin CLIEffort not specified15.5%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
Claude Opus 4.828 May 2026Claude Codemax effort15.3%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
GLM-5.3-Flash26 Aug 2026Devin CLImax effort14.9%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture
SWE-1.67 Apr 2026Devin CLIEffort not specified1.3%FrontierCode LeaderboardChecked A person checked this number against its source on 20 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Cognition’s leaderboard, 18 Nov 2026Our capture

All benchmarks