Grok 4.5

xAIReleased 8 Jul 2026ClosedFlagship

Compare Announcement

163152–174Index score#36 of 229 models

4/ 8Core tests taken4 of 6 areas · plus 15 anchor tests

−22Versus the frontier at releasebest then: Claude Fable 5Replaced by Grok 4.6

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

500KContext windowtokens it reads at once

Where it sitson the frontier

Claude Opus 4.6 reached this score 5 months earlier.

Test resultsby area · tick = best by any model

Reasoning & knowledgeevidence

GPQA Diamond

93%

Humanity's Last Exam

no result

Mathsevidence

FrontierMath T1-3

57%

Codingevidence

SWE-bench Verified

no result

Terminal-Bench

no result

Agentsevidence

APEX-Agents

56%

Novel problemsevidence

ARC-AGI-2

53%

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#36of 219 · 163

LMArena · Creative writingcreative writing#33of 386 · 1451 LMArena · Mathsmaths questions in chat#43of 373 · 1473 LMArena · Textoverall chat quality, judged by people#45of 388 · 1466 LMArena · Codingcoding questions in chat#45of 383 · 1514 Epoch Capabilities Indexoverall capability from many benchmarks#35of 239 · 154.0 LMArena · Visionunderstanding images#25of 149 · 1279 LMArena · Searchanswering with web search#8of 34 · 1213 LMArena · WebDevbuilding web apps and front ends#34of 123 · 1552 LiveBench · Overalloverall, with fresh questions to limit contamination#30of 58 · 75.8 LMArena · Documentsworking with long documents#24of 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.