CursorBench 4.0
Gemini 4 Carbon 60.4% leads since October 2026 Ambiguous multi-file coding tasks from real Cursor sessions in Cursor's production harness. Agentic graders score correctness, with cost reported.
Added long-horizon problems: investigation, intent understanding, managing jobs and design adherence. Scores do not compare with 3.x.
- Made by
- Cursor
- Released
- September 2026
- Kind
- Fixing real repositories
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Code
- Tasks
- Kept private
- Strongest model tested
- Claude Fable 5.2 #1 on the capability index
- Last in a launch post
- September 2026
- Last model added
- November 2026
- Scores refreshed
- November 13th, 2026
21
Models scored on its own leaderboard
3
Launch posts list it, from 2 labs
4
System cards report it, from 2 labs
Gemini 4 Carbon leads with 60.4%
Best score per model from the benchmark's own leaderboard, read on November 13th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
- 60.4 Gemini 4 Carbon
- 60.0 Claude Fable 5.2
- 57.8 Claude Opus 5.5
- 55.5 Claude Sonnet 5.5
- 54.3 Gemini 4 Hydrogen
- 52.9 Gemini 4 Argon
- 51.8 Claude Fable 5.1
- 50.4
Muse Wonder - 48.4 Claude Haiku 5.5
- 47.7
DeepSeek V4.1 Flash 1017 - 46.6 Claude Opus 5
- 46.3 Grok 4.7
- 42.6 GLM-5.3
- 41.7 GPT-5.6 Sol
- 41.6 Muse Spark 1.3
- 41.4 Grok 4.6
- 41.3 GPT-5.6 Terra
- 36.8 GLM-5.3-Flash
- 35.9 GPT-5.6 Luna
- 34.1 Claude Sonnet 5
- 27.7 Composer 2.5
- 1 Gemini 4 Carbon high 60.4%
- 2 Claude Fable 5.2 max 60%
- 3 Claude Opus 5.5 max 57.8%
- 4 Claude Sonnet 5.5 max 55.5%
- 5 Gemini 4 Hydrogen high 54.3%
- 6 Gemini 4 Argon high 52.9%
- 7 Claude Fable 5.1 max 51.8%
- 8 Muse Wonder max 50.4%
- 9 Claude Haiku 5.5 max 48.4%
- 10 DeepSeek V4.1 Flash 1017 max 47.7%
- 11 Claude Opus 5 max 46.6%
- 12 Grok 4.7 xhigh 46.3%
- 13 GLM-5.3 max 42.6%
- 14 GPT-5.6 Sol max 41.7%
- 15 Muse Spark 1.3 max 41.6%
- 16 Grok 4.6 xhigh 41.4%
- 17 GPT-5.6 Terra max 41.3%
- 18 GLM-5.3-Flash max 36.8%
- 19 GPT-5.6 Luna max 35.9%
- 20 Claude Sonnet 5 max 34.1%
- 21 Composer 2.5 27.7%
Source: Official CursorBench 4.0 leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon | 60.4% | high, Cursor | Oct 2026 | Official leaderboard |
| 2 | Claude Fable 5.2 | 60% | max, Cursor | Nov 2026 | Official leaderboard |
| 3 | Claude Opus 5.5 | 57.8% | max, Cursor | Sep 2026 | Official leaderboard |
| 4 | Claude Sonnet 5.5 | 55.5% | max, Cursor | Sep 2026 | Official leaderboard |
| 5 | Gemini 4 Hydrogen | 54.3% | high, Cursor | Nov 2026 | Official leaderboard |
| 6 | Gemini 4 Argon | 52.9% | high, Cursor | Sep 2026 | Official leaderboard |
| 7 | Claude Fable 5.1 | 51.8% | max, Cursor | Sep 2026 | Official leaderboard |
| 8 | Muse Wonder | 50.4% | max, Cursor | Oct 2026 | Official leaderboard |
| 9 | Claude Haiku 5.5 | 48.4% | max, Cursor | Oct 2026 | Official leaderboard |
| 10 | DeepSeek V4.1 Flash 1017 | 47.7% | max, Cursor | Oct 2026 | Official leaderboard |
| 11 | Claude Opus 5 | 46.6% | max, Cursor | Jul 2026 | Official leaderboard |
| 12 | Grok 4.7 | 46.3% | xhigh, Cursor | Sep 2026 | Official leaderboard |
| 13 | GLM-5.3 | 42.6% | max, Cursor | Aug 2026 | Official leaderboard |
| 14 | GPT-5.6 Sol | 41.7% | max, Cursor | Jul 2026 | Official leaderboard |
| 15 | Muse Spark 1.3 | 41.6% | max, Cursor | Sep 2026 | Official leaderboard |
| 16 | Grok 4.6 | 41.4% | xhigh, Cursor | Aug 2026 | Official leaderboard |
| 17 | GPT-5.6 Terra | 41.3% | max, Cursor | Jul 2026 | Official leaderboard |
| 18 | GLM-5.3-Flash | 36.8% | max, Cursor | Aug 2026 | Official leaderboard |
| 19 | GPT-5.6 Luna | 35.9% | max, Cursor | Jul 2026 | Official leaderboard |
| 20 | Claude Sonnet 5 | 34.1% | max, Cursor | Jun 2026 | Official leaderboard |
| 21 | Composer 2.5 | 27.7% | Cursor | May 2026 | Official leaderboard |
Labs report scores for 4 models in their cards. We plot every source in the charts below.
CursorBench score by average cost per task
The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.
| Model | Setting | CursorBench score | Average cost per task |
|---|---|---|---|
| Gemini 4 Carbon | high, Cursor | 60.4% | $9.22 |
| Gemini 4 Carbon | medium, Cursor | 60.1% | $8.49 |
| Claude Fable 5.2 | max, Cursor | 60% | $32.79 |
| Gemini 4 Carbon | low, Cursor | 58.6% | $11.89 |
| Claude Opus 5.5 | max, Cursor | 57.8% | $13.43 |
| Claude Fable 5.2 | high, Cursor | 56.9% | $9.7 |
| Claude Fable 5.2 | xhigh, Cursor | 56.3% | $15.48 |
| Claude Opus 5.5 | high, Cursor | 56% | $3.97 |
| Claude Opus 5.5 | xhigh, Cursor | 56% | $6.98 |
| Claude Sonnet 5.5 | max, Cursor | 55.5% | $7.05 |
| Claude Fable 5.2 | medium, Cursor | 54.8% | $4.66 |
| Gemini 4 Hydrogen | high, Cursor | 54.3% | $5.59 |
| Claude Fable 5.2 | low, Cursor | 53.2% | $3.94 |
| Claude Sonnet 5.5 | xhigh, Cursor | 53.1% | $2.81 |
| Gemini 4 Argon | high, Cursor | 52.9% | $4.57 |
| Gemini 4 Argon | medium, Cursor | 52.8% | $4.39 |
| Claude Opus 5.5 | medium, Cursor | 52.5% | $2.91 |
| Claude Fable 5.1 | max, Cursor | 51.8% | $17.28 |
| Claude Fable 5.1 | xhigh, Cursor | 51.6% | $13.01 |
| Gemini 4 Argon | low, Cursor | 50.6% | $4.09 |
| Muse Wonder | max, Cursor | 50.4% | $5.25 |
| Claude Fable 5.1 | high, Cursor | 49.2% | $9.08 |
| Claude Haiku 5.5 | max, Cursor | 48.4% | $1.12 |
| Muse Wonder | xhigh, Cursor | 48.2% | $4.24 |
| Claude Sonnet 5.5 | high, Cursor | 47.8% | $1.2 |
| DeepSeek V4.1 Flash 1017 | max, Cursor | 47.7% | $0.98 |
| Claude Fable 5.1 | medium, Cursor | 46.8% | $7.05 |
| Gemini 4 Hydrogen | medium, Cursor | 46.7% | $1.76 |
| Claude Opus 5 | max, Cursor | 46.6% | $11.95 |
| DeepSeek V4.1 Flash 1017 | medium, Cursor | 46.3% | $0.8 |
| Grok 4.7 | xhigh, Cursor | 46.3% | $6.01 |
| Muse Wonder | high, Cursor | 46.1% | $3.51 |
| Claude Opus 5 | xhigh, Cursor | 46.1% | $11.43 |
| DeepSeek V4.1 Flash 1017 | high, Cursor | 45.6% | $0.77 |
| Muse Wonder | medium, Cursor | 45.2% | $2.82 |
| Claude Fable 5.1 | low, Cursor | 45.1% | $5.44 |
| DeepSeek V4.1 Flash 1017 | low, Cursor | 44.9% | $0.6 |
| Claude Opus 5 | high, Cursor | 44.7% | $9 |
| Claude Haiku 5.5 | xhigh, Cursor | 44.3% | $0.56 |
| Grok 4.7 | high, Cursor | 43.9% | $4.69 |
| Claude Opus 5.5 | low, Cursor | 43.7% | $1.17 |
| Claude Opus 5 | medium, Cursor | 43.3% | $6.94 |
| GLM-5.3 | max, Cursor | 42.6% | $5.05 |
| Claude Haiku 5.5 | high, Cursor | 42.3% | $0.32 |
| Muse Wonder | low, Cursor | 41.7% | $1.91 |
| GPT-5.6 Sol | max, Cursor | 41.7% | $8.23 |
| Muse Spark 1.3 | max, Cursor | 41.6% | $2.64 |
| Grok 4.7 | medium, Cursor | 41.6% | $3.49 |
| Grok 4.6 | xhigh, Cursor | 41.4% | $6.1 |
| GPT-5.6 Terra | max, Cursor | 41.3% | $5.14 |
| Claude Opus 5 | low, Cursor | 40.7% | $4.87 |
| Grok 4.6 | high, Cursor | 40.4% | $5.2 |
| Claude Sonnet 5.5 | medium, Cursor | 39.2% | $0.52 |
| GLM-5.3 | high, Cursor | 38% | $3.24 |
| GPT-5.6 Sol | xhigh, Cursor | 37.7% | $4.4 |
| Muse Spark 1.3 | xhigh, Cursor | 37.5% | $2.1 |
| Claude Haiku 5.5 | medium, Cursor | 36.9% | $0.17 |
| GLM-5.3-Flash | max, Cursor | 36.8% | $0.39 |
| Grok 4.6 | medium, Cursor | 36.1% | $3.48 |
| GPT-5.6 Luna | max, Cursor | 35.9% | $1.03 |
| Claude Sonnet 5.5 | low, Cursor | 35.8% | $0.37 |
| GPT-5.6 Sol | high, Cursor | 35.7% | $2.85 |
| Muse Wonder | minimal, Cursor | 34.2% | $0.84 |
| Claude Sonnet 5 | max, Cursor | 34.1% | $7.17 |
| GPT-5.6 Terra | xhigh, Cursor | 33.6% | $1.81 |
| Muse Spark 1.3 | high, Cursor | 33.4% | $1.66 |
| Grok 4.6 | low, Cursor | 33.4% | $2.25 |
| GLM-5.3 | low, Cursor | 33.3% | $2.04 |
| Grok 4.7 | low, Cursor | 33.1% | $1.58 |
| GPT-5.6 Luna | xhigh, Cursor | 33% | $0.44 |
| Muse Spark 1.3 | medium, Cursor | 32.6% | $1.49 |
| Claude Sonnet 5 | xhigh, Cursor | 32% | $4.55 |
| GLM-5.3-Flash | high, Cursor | 31.1% | $0.25 |
| GPT-5.6 Sol | medium, Cursor | 31.1% | $1.77 |
| Claude Haiku 5.5 | low, Cursor | 30.9% | $0.08 |
| Claude Sonnet 5 | high, Cursor | 30.8% | $3.48 |
| GPT-5.6 Terra | high, Cursor | 30.7% | $1.11 |
| GPT-5.6 Luna | high, Cursor | 29.4% | $0.25 |
| Muse Spark 1.3 | low, Cursor | 29.3% | $0.93 |
| Claude Sonnet 5 | medium, Cursor | 28% | $2.31 |
| Gemini 4 Hydrogen | low, Cursor | 27.9% | $0.28 |
| Composer 2.5 | Cursor | 27.7% | $0.68 |
| GPT-5.6 Terra | medium, Cursor | 27.6% | $0.64 |
| GLM-5.3-Flash | low, Cursor | 26.9% | $0.15 |
| GPT-5.6 Terra | low, Cursor | 25.2% | $0.52 |
| GPT-5.6 Sol | low, Cursor | 24.6% | $0.87 |
| Muse Spark 1.3 | minimal, Cursor | 24.3% | $0.56 |
| Claude Sonnet 5 | low, Cursor | 24.1% | $1.39 |
| GPT-5.6 Luna | medium, Cursor | 22.2% | $0.08 |
| GPT-5.6 Luna | low, Cursor | 16% | $0.03 |
The best score, 60.4%, costs $9.22 a task
Runs on the cost frontier score higher than any run that's cheaper. Claude Opus 5.5 at high comes within 4.4 points of the best for $3.97 a task, about half the cost. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
OpenAI Anthropic DeepSeek Google DeepMind
Cost frontier
One model at each effort level
Source: Official CursorBench 4.0 leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Gemini 4 Carbon | high | Cursor | 60.4% | $9.22 | 39,960 |
| Gemini 4 Carbon | medium | Cursor | 60.1% | $8.49 | 37,067 |
| Claude Fable 5.2 | max | Cursor | 60% | $33 | 352,794 |
| Gemini 4 Carbon | low | Cursor | 58.6% | $12 | 42,663 |
| Claude Opus 5.5 | max | Cursor | 57.8% | $13 | 218,363 |
| Claude Fable 5.2 | high | Cursor | 56.9% | $9.70 | 99,259 |
| Claude Fable 5.2 | xhigh | Cursor | 56.3% | $15 | 161,251 |
| Claude Opus 5.5 | xhigh | Cursor | 56% | $6.98 | 101,083 |
| Claude Opus 5.5 | high | Cursor | 56% | $3.97 | 53,078 |
| Claude Sonnet 5.5 | max | Cursor | 55.5% | $7.05 | 271,920 |
| Claude Fable 5.2 | medium | Cursor | 54.8% | $4.66 | 44,103 |
| Gemini 4 Hydrogen | high | Cursor | 54.3% | $5.59 | 88,682 |
| Claude Fable 5.2 | low | Cursor | 53.2% | $3.94 | 36,616 |
| Claude Sonnet 5.5 | xhigh | Cursor | 53.1% | $2.81 | 100,158 |
| Gemini 4 Argon | high | Cursor | 52.9% | $4.57 | 109,500 |
| Gemini 4 Argon | medium | Cursor | 52.8% | $4.39 | 103,600 |
| Claude Opus 5.5 | medium | Cursor | 52.5% | $2.91 | 37,954 |
| Claude Fable 5.1 | max | Cursor | 51.8% | $17 | 117,236 |
| Claude Fable 5.1 | xhigh | Cursor | 51.6% | $13 | 87,294 |
| Gemini 4 Argon | low | Cursor | 50.6% | $4.09 | 94,200 |
| Muse Wonder | max | Cursor | 50.4% | $5.25 | 49,692 |
| Claude Fable 5.1 | high | Cursor | 49.2% | $9.08 | 58,438 |
| Claude Haiku 5.5 | max | Cursor | 48.4% | $1.12 | 325,934 |
| Muse Wonder | xhigh | Cursor | 48.2% | $4.24 | 39,564 |
| Claude Sonnet 5.5 | high | Cursor | 47.8% | $1.20 | 37,391 |
| DeepSeek V4.1 Flash 1017 | max | Cursor | 47.7% | $0.98 | 537,047 |
| Claude Fable 5.1 | medium | Cursor | 46.8% | $7.05 | 45,411 |
| Gemini 4 Hydrogen | medium | Cursor | 46.7% | $1.76 | 36,781 |
| Claude Opus 5 | max | Cursor | 46.6% | $12 | 85,384 |
| Grok 4.7 | xhigh | Cursor | 46.3% | $6.01 | 70,141 |
| DeepSeek V4.1 Flash 1017 | medium | Cursor | 46.3% | $0.80 | 398,255 |
| Claude Opus 5 | xhigh | Cursor | 46.1% | $11 | 80,094 |
| Muse Wonder | high | Cursor | 46.1% | $3.51 | 33,612 |
| DeepSeek V4.1 Flash 1017 | high | Cursor | 45.6% | $0.77 | 385,196 |
| Muse Wonder | medium | Cursor | 45.2% | $2.82 | 27,124 |
| Claude Fable 5.1 | low | Cursor | 45.1% | $5.44 | 34,795 |
| DeepSeek V4.1 Flash 1017 | low | Cursor | 44.9% | $0.60 | 290,651 |
| Claude Opus 5 | high | Cursor | 44.7% | $9.00 | 61,405 |
| Claude Haiku 5.5 | xhigh | Cursor | 44.3% | $0.56 | 143,813 |
| Grok 4.7 | high | Cursor | 43.9% | $4.69 | 56,382 |
| Claude Opus 5.5 | low | Cursor | 43.7% | $1.17 | 15,811 |
| Claude Opus 5 | medium | Cursor | 43.3% | $6.94 | 45,272 |
| GLM-5.3 | max | Cursor | 42.6% | $5.05 | 96,387 |
| Claude Haiku 5.5 | high | Cursor | 42.3% | $0.32 | 77,057 |
| GPT-5.6 Sol | max | Cursor | 41.7% | $8.23 | 42,944 |
| Muse Wonder | low | Cursor | 41.7% | $1.91 | 18,880 |
| Muse Spark 1.3 | max | Cursor | 41.6% | $2.64 | 52,005 |
| Grok 4.7 | medium | Cursor | 41.6% | $3.49 | 36,683 |
| Grok 4.6 | xhigh | Cursor | 41.4% | $6.10 | 49,814 |
| GPT-5.6 Terra | max | Cursor | 41.3% | $5.14 | 60,814 |
| Claude Opus 5 | low | Cursor | 40.7% | $4.87 | 31,995 |
| Grok 4.6 | high | Cursor | 40.4% | $5.20 | 41,387 |
| Claude Sonnet 5.5 | medium | Cursor | 39.2% | $0.52 | 16,036 |
| GLM-5.3 | high | Cursor | 38% | $3.24 | 60,031 |
| GPT-5.6 Sol | xhigh | Cursor | 37.7% | $4.40 | 24,729 |
| Muse Spark 1.3 | xhigh | Cursor | 37.5% | $2.10 | 40,891 |
| Claude Haiku 5.5 | medium | Cursor | 36.9% | $0.17 | 42,659 |
| GLM-5.3-Flash | max | Cursor | 36.8% | $0.39 | 56,410 |
| Grok 4.6 | medium | Cursor | 36.1% | $3.48 | 24,893 |
| GPT-5.6 Luna | max | Cursor | 35.9% | $1.03 | 87,284 |
| Claude Sonnet 5.5 | low | Cursor | 35.8% | $0.37 | 11,668 |
| GPT-5.6 Sol | high | Cursor | 35.7% | $2.85 | 16,174 |
| Muse Wonder | minimal | Cursor | 34.2% | $0.84 | 9,761 |
| Claude Sonnet 5 | max | Cursor | 34.1% | $7.17 | 149,257 |
| GPT-5.6 Terra | xhigh | Cursor | 33.6% | $1.81 | 23,436 |
| Grok 4.6 | low | Cursor | 33.4% | $2.25 | 16,307 |
| Muse Spark 1.3 | high | Cursor | 33.4% | $1.66 | 30,654 |
| GLM-5.3 | low | Cursor | 33.3% | $2.04 | 31,983 |
| Grok 4.7 | low | Cursor | 33.1% | $1.58 | 15,677 |
| GPT-5.6 Luna | xhigh | Cursor | 33% | $0.44 | 40,598 |
| Muse Spark 1.3 | medium | Cursor | 32.6% | $1.49 | 27,255 |
| Claude Sonnet 5 | xhigh | Cursor | 32% | $4.55 | 83,373 |
| GPT-5.6 Sol | medium | Cursor | 31.1% | $1.77 | 10,111 |
| GLM-5.3-Flash | high | Cursor | 31.1% | $0.25 | 35,104 |
| Claude Haiku 5.5 | low | Cursor | 30.9% | $0.08 | 23,382 |
| Claude Sonnet 5 | high | Cursor | 30.8% | $3.48 | 61,146 |
| GPT-5.6 Terra | high | Cursor | 30.7% | $1.11 | 13,162 |
| GPT-5.6 Luna | high | Cursor | 29.4% | $0.25 | 23,368 |
| Muse Spark 1.3 | low | Cursor | 29.3% | $0.93 | 17,483 |
| Claude Sonnet 5 | medium | Cursor | 28% | $2.31 | 39,114 |
| Gemini 4 Hydrogen | low | Cursor | 27.9% | $0.28 | 8,975 |
| Composer 2.5 | Cursor | 27.7% | $0.68 | 17,347 | |
| GPT-5.6 Terra | medium | Cursor | 27.6% | $0.64 | 7,307 |
| GLM-5.3-Flash | low | Cursor | 26.9% | $0.15 | 17,831 |
| GPT-5.6 Terra | low | Cursor | 25.2% | $0.52 | 5,914 |
| GPT-5.6 Sol | low | Cursor | 24.6% | $0.87 | 4,885 |
| Muse Spark 1.3 | minimal | Cursor | 24.3% | $0.56 | 10,620 |
| Claude Sonnet 5 | low | Cursor | 24.1% | $1.39 | 23,772 |
| GPT-5.6 Luna | medium | Cursor | 22.2% | $0.08 | 7,642 |
| GPT-5.6 Luna | low | Cursor | 16% | $0.03 | 3,288 |
3 models have held number one on the official leaderboard since its release in September 2026
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Anthropic Google DeepMind xAI
Sources: Official CursorBench 4.0 leaderboard and labs’ model cards
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 Reported by the lab in its model card
Best so far
Sources: Official CursorBench 4.0 leaderboard and labs’ model cards
2 labs 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.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Sep 28, 2026 | Claude Sonnet 5.5 | 55.5 | Agentic coding CursorBench 4.0 |
| Sep 22, 2026 | Claude Opus 5.5 | 57.8 | Agentic coding CursorBench 4.0 |
| Sep 21, 2026 | Grok 4.7 | 46.3 | CursorBench 4.0 (Software engineering) |
4 system cards report it, from 2 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 9, 2026 | Claude Fable 5.2 system card | 60 | CursorBench 4.0, max effort, independently run by Cursor. Additional prompt and harness improvements may prevent comparisons with earlier system cards (Section 8.8) | Capabilities |
| Sep 28, 2026 | Claude Sonnet 5.5 system card | 55.5 | CursorBench 4.0 at max effort, measured by Cursor (8.8, Figure 8.8.A) | Capabilities |
| Sep 22, 2026 | Claude Opus 5.5 system card | 57.8 | CursorBench 4.0 at max effort, measured by Cursor (56.0% at xhigh and high, 52.5% at medium) (8.8, Figure 8.8.A) | Capabilities |
| Sep 21, 2026 | Grok 4.7 model card | 46.3 | CursorBench 4.0, xhigh, value from text (2.1). High = 43.9%. Charts have no printed values | Capabilities |
4 lab claims 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. 4 of them match, which usually means one side copied the other. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| Grok 4.7 | 46.3% CursorBench 4.0 (Software engineering) | 46.3% Leaderboard, xhigh, Cursor | Same |
| Claude Opus 5.5 | 57.8% Agentic coding CursorBench 4.0 | 57.8% Leaderboard, max, Cursor | Same |
| Claude Sonnet 5.5 | 55.5% Agentic coding CursorBench 4.0 | 55.5% Leaderboard, max, Cursor | Same |
| Claude Fable 5.2 | 60% CursorBench 4.0, max effort, independently run by Cursor. Additional prompt and harness improvements may prevent comparisons with earlier system cards (Section 8.8) | 60% Leaderboard, max, Cursor | Same |
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.
Cursor Anthropic OpenAI Google DeepMind xAI
Sources: Official leaderboards, boards copied by Epoch AI (CC BY 4.0) and labs’ model cards
More 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