SpaceXAIReasoning modelCompare with another model

Grok 4.7

Reasoning effort

Grok 4.7 is a SpaceXAI proprietary reasoning model, released 21 Sept 2026. Its best configuration (high reasoning effort) ranks #56 of 775 on the BenchLeader Index at 64.0 ±7.3, in the upper half. The ± is the point: 199 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in composite (82) and lowest in reasoning (61). At $3.00 per million tokens blended it is pricier than most ranked models. Output speed of 87 tokens per second puts it faster than most, with a first answer in 33.0 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 21 Sept 2026.

Blended price: $3.00/M; $2.00 in · $6.00 out
Output speed: 87 tok/s; measured by Artificial Analysis
First answer: 33 s; first token 1.91 s
Context: 500k
Full answer: 39 s; median, reasoning included
Cost per run: $2.73; one full Intelligence Index run
Released: 21 Sept 2026

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

Overall index64
Reasoning61
Coding65
Knowledge75
Long context64
Composite82

Versions

SpaceXAI has shipped 10 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
Grok 4.7highthis page	21 Sept 2026	64.0	#56
Grok 4.6medium	11 Aug 2026	64.2	#55
Grok 4.5high	8 Jul 2026	59.3	#133
Grok 4.3medium	17 Apr 2026	58.7	#147
Grok 4.2high	10 Mar 2026	–	–
Grok 4.20thinking	17 Feb 2026	58.7	#144
Grok 4.1	17 Nov 2025	57.4	#179
Grok 4	9 Jul 2025	58.0	#166
Grok 3	17 Feb 2025	51.5	#339
Grok 2	12 Dec 2024	41.9	#601

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. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
low	61.1	#103	77 tok/s	7.42 s	$0.0026	Coding 61 · Composite 77 · Knowledge 75 · Long context 65 · Multimodal 61 · Reasoning 50
medium	–	–	72 tok/s	1.91 s	$0.0026	Reasoning 56
highbest	64.0	#56	87 tok/s	33 s	$0.0026	Coding 65 · Composite 82 · Knowledge 75 · Long context 64 · Reasoning 61
xhigh	61.8	#94	91 tok/s	47 s	$0.0026	Agents & tools 58 · Coding 63 · Composite 82 · Human preference 63 · Knowledge 65 · Long context 63 · Maths 56 · Multimodal 56 · Reasoning 63
not stated	–	–	72 tok/s	1.91 s	$0.0026	Agents & tools 55 · Coding 62 · Composite 57 · Maths 43

What more thinking costs

Turning the effort up buys index points and multiplies the bill. Here xhigh costs 3.0× low for +0.7 on the index — 0.4 points per doubling of spend. Compare that with other models.

How its index has moved

21 Sept 2026 to 20 Oct 2026

high
xhigh
low

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

See all LLM benchmarks. 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	xhigh	not stated	Source
GPQA Diamond	–	–	–	92.7%#24	–	Epoch AI Benchmarking Hub
Arena Hard Prompts	–	–	–	1465#85	–	Arena (LMArena)
LiveBench 2.0 Reasoningnot in indexeffort not stated	–	–	–	–	82.7%#35	LiveBench
LiveBench 1.0 Reasoningnot in index	–	–	–	82.7%#47	–	LiveBench
Humanity's Last Exam (AA)not in index	39.2%#92	–	42.3%#69	43.1%#59	–	Artificial Analysis
ARC-AGI-1	63.3%#147	90.0%#74	90.2%#73	89.5%#77	–	ARC Prize
ARC-AGI-2	16.7%#139	58.8%#88	58.3%#89	61.4%#75	–	ARC Prize
ARC-AGI-3	0.8%#38	8.4%#22	10.0%#20	7.0%#24	–	ARC Prize
CritPt	13.4%#100	–	18.0%#70	17.7%#73	–	Artificial Analysis
MysteryMechanismnot in index	–	–	–	25.2%#16	–	Vals AI
Chess Puzzlesnot in index	–	–	–	41.0%#39	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	–	–	–	29.0%#35	–	Epoch AI Benchmarking Hub
LMCAnot in index	–	–	–	49.4%#69	–	Epoch AI Benchmarking Hub
DTBenchnot in index	–	–	–	96.0%#30	–	Epoch AI Benchmarking Hub

Coding

Benchmark	low	medium	high	xhigh	not stated	Source
SciCode	54.9%#57	–	57.8%#24	57.4%#28	–	SciCode
FrontierCodeeffort not stated	–	–	–	–	47.6%#11	Cognition
Arena Coding	–	–	–	1497#85	–	Arena (LMArena)
Arena WebDev	–	–	–	1637#17	–	Arena (LMArena)
LiveBench 2.0 Codingnot in indexeffort not stated	–	–	–	–	84.5%#23	LiveBench
LiveBench 1.0 Codingnot in index	–	–	–	77.2%#39	–	LiveBench
SciCode (AA)not in index	54.9%#58	–	57.8%#24	57.4%#27	–	Artificial Analysis
IOI	–	–	–	57.7%#17	–	Vals AI
Code Migrationnot in index	–	–	–	44.8%#18	–	Vals AI
ProgramBenchnot in index	–	–	–	0.5%#18	–	Vals AI
Vibe Code Bench v1.1not in index	–	–	–	86.2%#12	–	Vals AI
CursorBenchnot in index	33.1%#58	41.6%#34	43.9%#27	46.3%#20	–	Cursor
FrontierSWEnot in index	–	–	–	29.5%#11	–	Epoch AI Benchmarking Hub
Terminal-Bench 4.0 (AA)not in index	16.2%#68	–	24.8%#55	25.8%#52	–	Artificial Analysis
Startrise Frontendnot in indexeffort not stated	–	–	–	–	75.9%#7	Startrise
FrontierCode 1.1 (AA)not in index	39.6%#27	–	34.1%#34	33.8%#35	–	Artificial Analysis

Agents & tools

Benchmark	low	high	xhigh	not stated	Source
Terminal-Bench	–	–	37.6%#49	–	Terminal-Bench
APEX-Agentseffort not stated	–	–	–	54.6%#20	Mercor
Arena Agent	–	–	3.3#15	–	Arena (LMArena)
LiveBench Long Agenticnot in indexeffort not stated	–	–	–	50.1%#13	LiveBench
LiveBench Browsernot in indexeffort not stated	–	–	–	60.0%#28	LiveBench
LiveBench Agentic Codingnot in index	–	–	54.0%#34	–	LiveBench
Terminal-Bench 2.1 (Vals)	–	–	73.4%#22	–	Vals AI
GDPval-AA v2.1not in index	54.4%#36	60.3%#12	60.5%#11	–	Artificial Analysis
ITBench SRE (AA)not in index	–	–	42.1%#16	–	Artificial Analysis
Analyst Agent (AA)not in index	–	55.0%#5	52.5%#8	–	Artificial Analysis
CyberBenchnot in index	–	–	69.5%#25	–	Vals AI
Finance Agent v2not in index	–	–	52.3%#35	–	Vals AI
Harvey's Legal Agent Benchmarknot in index	–	–	18.3%#4	–	Vals AI
Legal Research Benchnot in index	–	–	47.1%#14	–	Vals AI
SREBenchnot in index	–	–	3.8%#19	–	Vals AI
Tax Agent Benchnot in index	–	–	34.2%#13	–	Vals AI
Terminal-Bench 4.0 (Vals)not in index	–	–	28.8%#20	–	Vals AI
Terminal-Bench Sciencenot in index	–	–	10.0%#15	–	Vals AI
Vending-Bench 2not in indexeffort not stated	–	–	–	10536.8#7	Andon Labs
AutomationBenchnot in index	56.7%#50	63.5%#22	65.6%#15	–	Artificial Analysis
GDP.pdf (AA)not in index	21.4%#53	23.2%#43	20.0%#64	–	Artificial Analysis
Harvey LAB-AA v1.1not in index	–	–	9.4%#3	–	Harvey
AA-Briefcase v1.1not in index	–	–	1642#11	–	Artificial Analysis
Terminal-Bench-Science 0.1 (AA)not in index	4.8%#38	9.5%#28	4.8%#38	–	Artificial Analysis
CWE-Bench (AA)	–	–	68.3%#1	–	Artificial Analysis
DeepSec-Bench (AA)	–	–	26.9%#4	–	Artificial Analysis
CyberGym end-to-end (AA)	–	–	74.0%#5	–	Artificial Analysis

Maths

Benchmark	xhigh	not stated	Source
FrontierMath Tiers 1–3	53.0%#51	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	17.1%#55	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	98.1%#27	–	Epoch AI Benchmarking Hub
ProofBencheffort not stated	–	26.0%#44	Vals AI
LiveBench Mathematicsnot in index	95.7%#16	–	LiveBench
Arena Maths	1443#86	–	Arena (LMArena)

Knowledge

Benchmark	low	high	xhigh	not stated	Source
SimpleQA Verified	–	–	56.0%#20	–	Epoch AI Benchmarking Hub
LiveBench 2.0 Data Analysisnot in indexeffort not stated	–	–	–	76.9%#20	LiveBench
LiveBench Researchnot in indexeffort not stated	–	–	–	86.4%#5	LiveBench
LiveBench 1.0 Data Analysisnot in index	–	–	76.9%#30	–	LiveBench
AA-Omniscience	29.6#33	30.9#30	32.0#27	–	Artificial Analysis
LegalBench	–	–	84.4%#33	–	Vals AI
PRBench Legal	–	–	47.6%#19	–	Scale AI SEAL
BioMysteryBenchnot in index	–	–	69.3%#13	–	Vals AI
Excel Modeling Benchmarknot in index	–	–	67.0%#19	–	Vals AI
MedCodenot in index	–	–	49.5%#22	–	Vals AI
MedScribenot in index	–	–	89.4%#7	–	Vals AI
Public Benefits Benchnot in index	–	–	65.6%#20	–	Vals AI
GDP.pdfnot in index	–	–	22.8%#26	–	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	49.7%#63	47.8%#73	47.5%#75	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	60.1%#76	67.6%#54	70.7%#39	–	Artificial Analysis

Instruction following

Benchmark	xhigh	Source
LiveBench Languagenot in index	80.1%#30	LiveBench
LiveBench Instruction Followingnot in index	75.3%#10	LiveBench
Arena Instruction Followingnot in index	1439#79	Arena (LMArena)

Human preference

Benchmark	xhigh	Source
Arena Text	1444#87	Arena (LMArena)
Arena Creative Writingnot in index	1424#77	Arena (LMArena)
Arena Multi-turnnot in index	1428#115	Arena (LMArena)
Arena Longer Queriesnot in index	1454#81	Arena (LMArena)

Multimodal

Benchmark	low	xhigh	not stated	Source
Arena Vision	–	1231#77	–	Arena (LMArena)
LiveBench Designnot in indexeffort not stated	–	–	78.0%#20	LiveBench
MMMU-Pro	78.2%#75	–	–	Artificial Analysis
SAGEnot in index	–	40.8%#58	–	Vals AI
Blueprint-Bench 2not in indexeffort not stated	–	–	32.5%#14	Epoch AI Benchmarking Hub
AA-3Dnot in index	–	995#11	–	Artificial Analysis

Long context

Benchmark	low	high	xhigh	Source	Trend
AA-LCR	80.0%#77	77.0%#137	76.7%#139	Artificial Analysis
MLCRnot in index	–	–	15.0%#46	Artificial Analysis

Composite

Benchmark	low	high	xhigh	not stated	Source
Epoch Capabilities Indexnot in indexeffort not stated	–	–	–	154.4#41	Epoch AI Benchmarking Hub
LiveBench 2.0effort not stated	–	–	–	74.1%#20	LiveBench
LiveBench 1.0not in index	–	–	77.4%#25	–	LiveBench
AA Intelligence Index v4.3.2	42.2#52	46.3#32	46.5#31	–	Artificial Analysis
Vals Indexnot in index	–	–	58.6#11	–	Vals AI
AA Cyber Indexnot in index	–	–	56.4#1	–	Artificial Analysis

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	83 tok/s	1.44 s	$4.00	$12.00	500k	–
SpaceXAI	82 tok/s	1.84 s	$2.00	$6.00	500k	–
SpaceXAI	77 tok/s	1.85 s	$2.00	$6.00	500k	–
SpaceXAI	71 tok/s	1.71 s	$2.20	$6.60	500k	–
SpaceXAI	63 tok/s	1.98 s	$4.00	$12.00	500k	–

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $2.00 in · $6.00 out

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 $0.500 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0026	$0.0022	36.5 s
Summarize a 30-page report	12,000 / 600	$0.028	$0.014	40.0 s
Code edit	6,000 / 1,500	$0.021	$0.014	50.4 s
Agentic coding session	60,000 / 4,000	$0.144	$0.076	1.3 min
Structured extraction	2,000 / 200	$0.0052	$0.0030	35.4 s