xAIReasoning modelAuto-detected

Grok 4.5

Name: Grok 4.5 benchmark results
Creator: BenchLeader

Best configuration ranks #47 of 610 on the BenchLeader Index at 63.4 ±3.8. Last measured 8 Sept 2026. Released 8 Jul 2026.

Blended price: $3.00/M; $2.00 in · $6.00 out
Output speed: 56 tok/s
First answer: 13 s; first token 1.08 s
Context: 500k

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index63
Reasoning69
Coding60
Agents & tools59
Knowledge76
Human preference67
Multimodal65
Long context66
Composite66

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	–	–	56 tok/s	13 s	$0.0026	Reasoning 49
medium	–	–	56 tok/s	13 s	$0.0026	Reasoning 55
high	54.2	#204	56 tok/s	13 s	$0.0026	Agents & tools 37 · Coding 63 · Knowledge 60 · Maths 55 · Reasoning 59
defaultbest	63.4	#47	56 tok/s	13 s	$0.0026	Agents & tools 59 · Coding 60 · Composite 66 · Human preference 67 · Knowledge 76 · Long context 66 · Multimodal 65 · Reasoning 69

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	low	medium	high	default	Source
GPQA Diamond	–	–	93.4%#13	–	Epoch AI Benchmarking Hub
SimpleBench	–	–	–	70.0%#14	SimpleBench
LMArena Hard Prompts	–	–	–	1495#33	LMArena
LiveBench Reasoningnot in index	–	–	–	87.2%#22	LiveBench
GPQA Diamond (AA)not in index	–	–	–	93.1%#24	Artificial Analysis
Humanity's Last Exam (AA)not in index	–	–	–	42.7%#36	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	92.9%#12	–	Vals AI
Kagi LLM Benchmark	–	–	–	83.5%#6	Kagi LLM Benchmark
ARC-AGI-1	79.2%#83	87.2%#61	85.7%#73	–	ARC Prize
ARC-AGI-2	33.1%#86	52.6%#71	52.6%#71	–	ARC Prize
ARC-AGI-3	0.3%#26	0.3%#26	0.3%#26	–	ARC Prize

Coding

Benchmark	low	medium	high	default	Source
SciCode	–	–	54.0%#34	–	SciCode
WeirdML	–	–	–	46.4%#76	WeirdML
FrontierCode	–	–	–	42.4%#12	Cognition
LMArena Coding	–	–	–	1523#22	LMArena
LMArena WebDev	–	–	–	1556#23	LMArena
LiveBench Codingnot in index	–	–	–	68.6%#49	LiveBench
SciCode (AA)not in index	–	–	–	55.0%#34	Artificial Analysis
LiveCodeBench	–	–	87.3%#14	–	Vals AI
SWE-bench (Vals)not in index	–	–	86.6%#13	–	Vals AI

Agents & tools

Benchmark	low	medium	high	default	Source
Terminal-Bench	–	–	12.4%#61	–	Terminal-Bench
APEX-Agents	–	–	–	34.2%#20	Mercor
LMArena Agent	–	–	–	3.9#14	LMArena
LiveBench Agentic Codingnot in index	–	–	–	56.5%#20	LiveBench
Terminal-Bench 2.1 (Vals)	–	–	67.8%#25	–	Vals AI

Maths

Benchmark	low	medium	high	default	Source
FrontierMath Tiers 1–3	–	–	57.2%#31	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	24.4%#36	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	–	–	97.8%#17	–	Epoch AI Benchmarking Hub
ProofBench	–	–	31.0%#28	–	Vals AI
LiveBench Mathematicsnot in index	–	–	–	90.8%#22	LiveBench

Knowledge

Benchmark	low	medium	high	default	Source
SimpleQA Verified	–	–	48.3%#26	–	Epoch AI Benchmarking Hub
LiveBench Data Analysisnot in index	–	–	–	73.0%#32	LiveBench
AA-Omniscience	–	–	–	25.3#30	Artificial Analysis
MMLU-Pro	–	–	89.2%#14	–	Vals AI
LegalBench	–	–	86.0%#16	–	Vals AI
CorpFin	–	–	67.4%#12	–	Vals AI
TaxEval	–	–	71.7%#73	–	Vals AI

Instruction following

Benchmark	low	medium	high	default	Source	Trend
LiveBench Languagenot in index	–	–	–	82.8%#16	LiveBench

Human preference

Benchmark	low	medium	high	default	Source	Trend
LMArena Text	–	–	–	1471#38	LMArena

Multimodal

Benchmark	low	medium	high	default	Source	Trend
LMArena Vision	–	–	–	1291#22	LMArena
MMMU-Pro	–	–	–	80.4%#34	Artificial Analysis

Long context

Benchmark	low	medium	high	default	Source	Trend
AA-LCR	–	–	–	79.3%#65	Artificial Analysis

Composite

Benchmark	low	medium	high	default	Source
Epoch Capabilities Indexnot in index	–	–	–	153.9#30	Epoch AI Benchmarking Hub
LiveBench	–	–	–	75.8%#26	LiveBench
AA Intelligence Index	–	–	–	39.1#41	Artificial Analysis
Vals Indexnot in index	–	–	51.5#27	–	Vals AI

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
SpaceXAI (ZDR)	54 tok/s	0.96 s	$2.00	$6.00	500k	–
SpaceXAI Priority (ZDR)	50 tok/s	1.44 s	$4.00	$12.00	500k	–
SpaceXAI	48 tok/s	1.08 s	$2.00	$6.00	500k	–

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.300 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0026	$0.0021	18.1 s
Summarise a 30-page report	12,000 / 600	$0.028	$0.012	23.5 s
Code edit	6,000 / 1,500	$0.021	$0.013	39.7 s
Agentic coding session	60,000 / 4,000	$0.144	$0.068	1.4 min
Structured extraction	2,000 / 200	$0.0052	$0.0027	16.3 s