Grok 4.6

xAIReleased 12 Aug 2026ClosedFlagshipDate verified

Compare Announcement

170159–181Index score#19 of 229 models

4/ 8Core tests taken4 of 6 areas · plus 16 anchor tests

−17Versus the frontier at releasebest then: Claude Opus 5Replaced by Grok 4.7

$2.0/ $6.0Price per million tokensinput / output, list price

500KContext windowtokens it reads at once

Where it sitson the frontier

GPT-5.3-Codex reached this score 6 months earlier.

Test resultsby area · tick = best by any model

Reasoning & knowledgeevidence

GPQA Diamond

94%

Humanity's Last Exam

no result

Mathsevidence

FrontierMath T1-3

66%

Codingevidence

SWE-bench Verified

no result

Terminal-Bench

no result

Agentsevidence

APEX-Agents

65%

Novel problemsevidence

ARC-AGI-2

67%

Long tasksno evidence yet

METR time horizon

no result

Independent results from the Epoch AI benchmarking hub, best reasoning setting. The core tests are shown here; anchor tests also inform the score and can be shown above.

What others sayrank on each leaderboard

The Model Indexindependent tests, our method#19of 219 · 170

Epoch Capabilities Indexoverall capability from many benchmarks#19of 239 · 156.6 LMArena · Creative writingcreative writing#45of 386 · 1445 LMArena · WebDevbuilding web apps and front ends#18of 123 · 1620 LMArena · Codingcoding questions in chat#58of 383 · 1507 LMArena · Textoverall chat quality, judged by people#66of 388 · 1453 LMArena · Mathsmaths questions in chat#73of 373 · 1446 LMArena · Visionunderstanding images#36of 149 · 1263 LiveBench · Overalloverall, with fresh questions to limit contamination#15of 58 · 78 LMArena · Documentsworking with long documents#23of 41 · 1452

Further left is better. Leaderboards measure different things: LMArena is people voting blind between two answers, the others are test scores. Click any row for the source.