# Benchmarks use the "Grok 4.6 High" comparison column and Grok 4.6 (high) GDPval chart in the Grok 4.7 announcement.
name = "Grok 4.6"
description = "xAI's frontier model for long-running agents, coding, knowledge work, and visual projects"
family = "grok"
knowledge = "2026-02-01"
release_date = "2026-08-12"
last_updated = "2026-08-12"
attachment = true
reasoning = true
temperature = true
tool_call = true
structured_output = true
open_weights = false

[limit]
context = 500_000
output = 500_000

[modalities]
input = ["text", "image"]
output = ["text"]

[[benchmarks]]
name = "CursorBench"
score = 40.4
metric = "score"
variant = "high effort"
version = "4.0"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "DeepSWE"
score = 65.2
metric = "score"
variant = "high effort"
version = "1.1"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "EEBench"
score = 53
metric = "score"
variant = "high effort"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "AA-Briefcase"
score = 1546
metric = "Elo"
variant = "high effort"
version = "1.1"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "Terminal-Bench"
score = 20.3
metric = "score"
variant = "high effort"
version = "4.0"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "Harvey Legal Agent Benchmark"
score = 15.8
metric = "score"
variant = "high effort"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "HealthBench"
score = 48.5
metric = "score"
variant = "high effort; professional"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"

[[benchmarks]]
name = "GDPval"
score = 1605
metric = "Elo"
variant = "high effort"
source = "https://x.ai/news/grok-4-7"
date = "2026-09-21"