Scores over time

Best recorded score per model; higher is better. The line follows the running best. Evaluation contexts can differ.

CursorBench Accuracy scores over timeOne point per model's best recorded score, dated by reported date or model release date. A line shows the running best. The data table below contains every plotted point.57.827.72026-09-102026-10-08Claude Sonnet 5: 34.1% · Reported 2026-09-10Claude Opus 5: 46.6% · Reported 2026-09-10Claude Fable 5.1: 51.8% · Reported 2026-09-10Composer 2.5: 27.7% · Reported 2026-09-10Grok 4.7: 46.3% · Reported 2026-09-21Claude Opus 5.5: 57.8% · Reported 2026-09-22Claude Sonnet 5.5: 55.5% · Reported 2026-09-28GLM-5.3-Flash: 36.8% · Reported 2026-10-07GPT-5.6 Luna: 35.9% · Reported 2026-10-07Gemini 3.8 Flash: 39.6% · Reported 2026-10-07Muse Spark 1.3: 41.6% · Reported 2026-10-07GLM-5.3: 42.6% · Reported 2026-10-07Grok 4.6: 41.4% · Reported 2026-10-07GPT-5.6 Sol: 41.7% · Reported 2026-10-07GPT-5.6 Terra: 41.3% · Reported 2026-10-07Claude Haiku 5.5: 48.4% · Reported 2026-10-08
Data table
Scores plotted over time
ModelDateScore
Claude Sonnet 5 (Reported)34.1%
Claude Opus 5 (Reported)46.6%
Claude Fable 5.1 (Reported)51.8%
Composer 2.5 (Reported)27.7%
Grok 4.7 (Reported)46.3%
Claude Opus 5.5 (Reported)57.8%
Claude Sonnet 5.5 (Reported)55.5%
GLM-5.3-Flash (Reported)36.8%
GPT-5.6 Luna (Reported)35.9%
Gemini 3.8 Flash (Reported)39.6%
Muse Spark 1.3 (Reported)41.6%
GLM-5.3 (Reported)42.6%
Grok 4.6 (Reported)41.4%
GPT-5.6 Sol (Reported)41.7%
GPT-5.6 Terra (Reported)41.3%
Claude Haiku 5.5 (Reported)48.4%

Results

67 results

CursorBench 4.0 results
ProviderModelScoreSourceRegistry No.
AnthropicClaude Haiku 5.5 (xhigh)44.3%cursor.com for Claude Haiku 5.5 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Haiku 5.5 on CursorBench 4.020018
AnthropicClaude Haiku 5.5 (medium)36.9%cursor.com for Claude Haiku 5.5 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Haiku 5.5 on CursorBench 4.020018
AnthropicClaude Haiku 5.5 (high)42.3%cursor.com for Claude Haiku 5.5 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Haiku 5.5 on CursorBench 4.020018
AnthropicClaude Haiku 5.5 (low)30.9%cursor.com for Claude Haiku 5.5 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Haiku 5.5 on CursorBench 4.020018
AnthropicClaude Haiku 5.5 (max)48.4%cursor.com for Claude Haiku 5.5 (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Haiku 5.5 on CursorBench 4.020018
OpenAIGPT-5.6 Luna (max)35.9%cursor.com for GPT-5.6 Luna (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Luna on CursorBench 4.010009
OpenAIGPT-5.6 Luna (low)16.0%cursor.com for GPT-5.6 Luna (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Luna on CursorBench 4.010009
OpenAIGPT-5.6 Luna (medium)22.2%cursor.com for GPT-5.6 Luna (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Luna on CursorBench 4.010009
OpenAIGPT-5.6 Luna (high)29.4%cursor.com for GPT-5.6 Luna (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Luna on CursorBench 4.010009
OpenAIGPT-5.6 Luna (xhigh)33.0%cursor.com for GPT-5.6 Luna (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Luna on CursorBench 4.010009
OpenAIGPT-5.6 Sol (medium)31.1%cursor.com for GPT-5.6 Sol (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Sol on CursorBench 4.010010
OpenAIGPT-5.6 Sol (low)24.6%cursor.com for GPT-5.6 Sol (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Sol on CursorBench 4.010010
OpenAIGPT-5.6 Sol (xhigh)37.7%cursor.com for GPT-5.6 Sol (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Sol on CursorBench 4.010010
OpenAIGPT-5.6 Sol (max)41.7%cursor.com for GPT-5.6 Sol (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Sol on CursorBench 4.010010
OpenAIGPT-5.6 Sol (high)35.7%cursor.com for GPT-5.6 Sol (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Sol on CursorBench 4.010010
OpenAIGPT-5.6 Terra (xhigh)33.6%cursor.com for GPT-5.6 Terra (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Terra on CursorBench 4.010011
OpenAIGPT-5.6 Terra (low)25.2%cursor.com for GPT-5.6 Terra (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Terra on CursorBench 4.010011
OpenAIGPT-5.6 Terra (high)30.7%cursor.com for GPT-5.6 Terra (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Terra on CursorBench 4.010011
OpenAIGPT-5.6 Terra (medium)27.6%cursor.com for GPT-5.6 Terra (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Terra on CursorBench 4.010011
OpenAIGPT-5.6 Terra (max)41.3%cursor.com for GPT-5.6 Terra (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GPT-5.6 Terra on CursorBench 4.010011
Z.aiGLM-5.3 (low)33.3%cursor.com for GLM-5.3 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3 on CursorBench 4.0150005
Z.aiGLM-5.3 (max)42.6%cursor.com for GLM-5.3 (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3 on CursorBench 4.0150005
Z.aiGLM-5.3 (high)38.0%cursor.com for GLM-5.3 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3 on CursorBench 4.0150005
Z.aiGLM-5.3-Flash (max)36.8%cursor.com for GLM-5.3-Flash (max) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3-Flash on CursorBench 4.0150006
Z.aiGLM-5.3-Flash (high)31.1%cursor.com for GLM-5.3-Flash (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3-Flash on CursorBench 4.0150006
Z.aiGLM-5.3-Flash (low)26.9%cursor.com for GLM-5.3-Flash (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for GLM-5.3-Flash on CursorBench 4.0150006
AnthropicClaude Fable 5.1 (xhigh)51.6%cursor.com for Claude Fable 5.1 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Fable 5.1 on CursorBench 4.020004
AnthropicClaude Fable 5.1 (low)45.1%cursor.com for Claude Fable 5.1 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Fable 5.1 on CursorBench 4.020004
AnthropicClaude Fable 5.1 (medium)46.8%cursor.com for Claude Fable 5.1 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Fable 5.1 on CursorBench 4.020004
AnthropicClaude Fable 5.1 (high)49.2%cursor.com for Claude Fable 5.1 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Fable 5.1 on CursorBench 4.020004
AnthropicClaude Sonnet 5 (high)30.8%cursor.com for Claude Sonnet 5 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Sonnet 5 on CursorBench 4.020013
AnthropicClaude Sonnet 5 (xhigh)32.0%cursor.com for Claude Sonnet 5 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Sonnet 5 on CursorBench 4.020013
AnthropicClaude Sonnet 5 (low)24.1%cursor.com for Claude Sonnet 5 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Sonnet 5 on CursorBench 4.020013
AnthropicClaude Sonnet 5 (medium)28.0%cursor.com for Claude Sonnet 5 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Sonnet 5 on CursorBench 4.020013
AnthropicClaude Opus 5 (high)44.7%cursor.com for Claude Opus 5 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5 on CursorBench 4.020014
AnthropicClaude Opus 5 (low)40.7%cursor.com for Claude Opus 5 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5 on CursorBench 4.020014
AnthropicClaude Opus 5 (xhigh)46.1%cursor.com for Claude Opus 5 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5 on CursorBench 4.020014
AnthropicClaude Opus 5 (medium)43.3%cursor.com for Claude Opus 5 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5 on CursorBench 4.020014
AnthropicClaude Opus 5.5 (high)56.0%cursor.com for Claude Opus 5.5 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5.5 on CursorBench 4.020015
AnthropicClaude Opus 5.5 (xhigh)56.0%cursor.com for Claude Opus 5.5 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5.5 on CursorBench 4.020015
AnthropicClaude Opus 5.5 (low)43.7%cursor.com for Claude Opus 5.5 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Claude Opus 5.5 on CursorBench 4.020015
Google (DeepMind)Gemini 3.8 Flash (high)39.6%cursor.com for Gemini 3.8 Flash (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Gemini 3.8 Flash on CursorBench 4.030004
Google (DeepMind)Gemini 3.8 Flash (medium)37.3%cursor.com for Gemini 3.8 Flash (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Gemini 3.8 Flash on CursorBench 4.030004
SpaceXAIGrok 4.6 (low)33.4%cursor.com for Grok 4.6 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.6 on CursorBench 4.040002
SpaceXAIGrok 4.6 (xhigh)41.4%cursor.com for Grok 4.6 (xhigh) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.6 on CursorBench 4.040002
SpaceXAIGrok 4.6 (medium)36.1%cursor.com for Grok 4.6 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.6 on CursorBench 4.040002
SpaceXAIGrok 4.7 (medium)41.6%cursor.com for Grok 4.7 (medium) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.7 on CursorBench 4.040003
SpaceXAIGrok 4.7 (high)43.9%cursor.com for Grok 4.7 (high) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.7 on CursorBench 4.040003
SpaceXAIGrok 4.7 (low)33.1%cursor.com for Grok 4.7 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Grok 4.7 on CursorBench 4.040003
Meta AI (originally Facebook AI Research)Muse Spark 1.3 (low)29.3%cursor.com for Muse Spark 1.3 (low) on CursorBench 4.0 (opens in a new tab)Archived copy (opens in a new tab)CursorBench AccuracyReported Benchmark leaderboardCursorIndependentRecord for Muse Spark 1.3 on CursorBench 4.080005