Grok 4.6

Aug 2026
Grok 4.6
Compare in
on
$1.48per task?
$2 / $6per 1M in / outcache $0.50
65 tok/sfirst token in 37 s?
500Kcontext

Rankingat high effort?

Full ranking →
1Fable 5.167.22Opus 5.566.23Fable 565.44GPT-6 Astra62.65GPT-6.1 Sol62.1
···
27Opus 4.848.828Grok 4.648.229Kimi K347.6

Strong at

S
Agents & work tasks#17 of 175Reasoning#31 of 334Science#3 of 312Knowledge#11 of 318
A
Coding#29 of 132Long context#38 of 273

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperClaude Opus 5.5medium+34%-10%$1.34No forced tool use+5BetterGPT-6.1 Sollow+22%-91%$0.13Remove unsupported parameters+5OpenDeepSeek V4 Pro 0813maxsame-55%$0.67No image input+5Lower effortSame modellow-10%-75%$0.38CheaperGPT-6 Astralow+23%-45%$0.82Monitoring can return HTTP 403+2

History

8 versions
304050Jul 25Oct 25Jan 26Apr 26Jul 26Oct 26Release dateAA Omniscience · Accuracy44.20 · Mar4.20 · Apr4.34.54.64.7
304050Jul 25Jan 26Jul 26Release dateAA Omniscience · Accuracy44.20 · Mar4.20 · Apr4.34.54.64.7

Benchmarks

in Chat & Knowledge
AA Intelligence Index 4.3.244.3#19
AA Omniscience Index30.5#11
AA Omniscience · Accuracy48.2#28
AA Omniscience · Hallucination Rate34.3#47
Humanity's Last Exam42.9#31
GPQA Diamond94.9#3
AA-LCR (long context) 1.180.3#38
CritPt (physics)17.1#39
GDP.pdf · AA all-pass17#34
MLCR-AA12.2#37
ARC-AGI-1 · public evaluation96.3#19
ARC-AGI-1 · semi-private87#30
ARC-AGI-2 · public evaluation68.3#18
ARC-AGI-2 · semi-private65.1#22
Arena Text · overall1,454#47
Arena Text · overall · full cohort1,427.8#67
Arena Text · Chinese1,475.1#60
Arena Text · English1,432.5#72
Arena Text · French1,456.9#45
Arena Text · German1,431.9#49
Arena Text · instruction following1,430.6#53
Arena Text · Japanese1,380.5#59
Arena Text · Korean1,395.8#47
Arena Text · non-English1,418.9#62
Arena Text · Polish1,429.5#49
Arena Text · Russian1,421.8#64
Arena Text · Spanish1,403.7#83
GPQA Diamond · Epoch94#10
OTIS Mock AIME 2024 202597.8#25
SimpleQA Verified v149.3#29
Chess Puzzles33#40

FAQ

How much does Grok 4.6 cost?›

$2 per 1M input tokens and $6 per 1M output tokens; cached input $0.50 per 1M. A typical benchmark task costs about $1.48 at high effort.

Is Grok 4.6 free to use?›

There is no free API; the cheapest paid price is $2 per 1M input tokens.

How good is Grok 4.6, and what is it best at?›

#31 of 334 in Chat & Knowledge, #29 of 132 in Coding, #8 of 133 in Agents & Tool Use, #35 of 195 in Writing & Language, #21 of 37 in Vision & Documents. Strongest at: Coding, Agents & work tasks, Reasoning, Science, Long context, Knowledge.

Is Grok 4.6 better than Grok 4.3?›

On AA Omniscience · Accuracy, Grok 4.6 scores 48.2 vs 34.8 — higher. It is 623% more expensive per task.

How can I use Grok 4.6?›

Through the API of Amazon Bedrock, xAI.

What is the context window of Grok 4.6?›

500K tokens, with up to 450K output tokens.

When was Grok 4.6 released? Is there a newer version?›

Released 2026-08-12. It has been superseded by Grok 4.7.

Is Grok 4.6 open source?›

No, Grok 4.6 is a closed model available through an API. The closest open-weights option here is DeepSeek V4 Pro 0813.

Can Grok 4.6 read images?›

It accepts text and images, PDF files as input.

What is Grok 4.6?›

Grok 4.6 is a closed language model from xAI, released Aug 2026, the successor to Grok 4.3.

What are the best alternatives to Grok 4.6?›

Claude Opus 5.5 (cheaper), GPT-6.1 Sol (better), DeepSeek V4 Pro 0813 (open weights).