Gemini 4 Carbon Tops CursorBench in Cursor

// 4h agoBENCHMARK RESULT

Gemini 4 Carbon Tops CursorBench in Cursor

Google’s Gemini 4 Carbon is now available in Cursor, where its high-effort mode leads CursorBench at 60.4%. The model targets long-horizon coding, reasoning, and cybersecurity tasks.

// ANALYSIS

Carbon’s narrow benchmark win is notable, but its real advantage may be delivering near-identical performance across effort levels.

–High effort scores 60.4%, barely ahead of medium at 60.1% and low at 58.6%
–It edges Fable 5.2 Max at 60.0% and Opus 5.5 Max at 57.8%
–Cursor reports $9.22 per benchmark task, so the lead comes with meaningful inference cost
–Developers should validate the result on their own repositories; benchmark rankings do not guarantee better latency, reliability, or code quality

// TAGS

gemini-4-carboncursorllmreasoningai-codingcoding-agentbenchmark

DISCOVERED

4h ago

2026-11-10

PUBLISHED

4h ago

2026-11-10

RELEVANCE

10/ 10

AUTHOR

cursor_ai

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

RESEARCH55m ago

Meta's AutoRefine Calibrates Code Judges

Meta’s AutoRefine paper presents a prompt-rewriting loop that uses noisy workflow outcomes—tests, merges, and abandons—to recalibrate LLM code judges, while abstaining when outcomes cannot be explained by the judge’s inputs. On held-out diffs from a deployed coding agent, agreement rose from Scott’s π 0.16 to 0.51. ([paper](https://arxiv.org/abs/2611.09291))

INFRA1h ago

Browser Use Cloud Clears CAPTCHA Bottleneck

Browser Use’s stealth cloud browser combines hardened Chromium, residential proxies, and automatic CAPTCHA solving for AI agents. The company reports 90.5% successful visits across 100 bot-protected sites in its own benchmark.

RESEARCH1h ago

Sycophantic AI Warnings Miss Their Mark

A preregistered study of 2,610 people found that warnings reduced perceived AI objectivity and trust, but did not reliably reduce users’ belief that they were right or their willingness to repair interpersonal conflicts. [Read the paper](https://arxiv.org/abs/2606.21317)