Inference
EN

Grok 4.6 vs Step 5 Preview

Change models
Grok 4.6xAI
Step 5 PreviewStepFun
IndexAll categories, out of 100.
67.8 ±3.7
68.4 ±3.7 best
GapIndex points below the top.
−0.6 · no clear gap
Top
CoverageIndex weight that has results.
85%
90% best
SpeedOutput tokens per second.
56/s
105/s best
EloLMArena rating and rank.
1430
1443 best
ReleasedMonth of release.
2026-08
2026-09
Input / 1MUS dollars per 1M input tokens.
$2.00
$1.00 best
Output / 1MUS dollars per 1M output tokens.
$6.00
$2.70 best
Batch output / 1MBatch price per 1M output tokens.
$4.80
—
1K chats2K tokens in, 500 out, each.
$7.00
$3.35 best
1K documents20K tokens in, 1K out, each.
$46.00
$22.70 best
1K agent tasks200K tokens in, 20K out, each.
$520.00
$254.00 best
ContextMaximum tokens in one request.
500K
1M best
AgenticMulti-step tasks with tools.
68.4
68.8 best
CodingCode writing and repair.
67.7
71.4 best
ReasoningLogic problems and puzzles.
66.1
67.4 best
MultimodalTasks with images and text.
—
62.2
KnowledgeFacts and expert knowledge.
65.7 best
63.5
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
72.0
—
MathMath problems.
65.3 best
61.7

Benchmarks

32 shared
Only Grok 4.6: 52Only Step 5 Preview: 30
Grok 4.6
Step 5 Preview
Artificial Analysis Humanity's Last ExamKnowledge
42.9%34 / 222
46.5% best23 / 222
Artificial Analysis Long Context ReasoningReasoning
80.3%40 / 217
88.3% best2 / 217
Critical Physics TasksReasoning
17.1%42 / 214
20.9% best26 / 214
Artificial Analysis Omniscience AccuracyKnowledge
48.2% best27 / 164
36.4%64 / 164
GDPval-AA normalizedAgentic
55.9% best16 / 133
54.3%24 / 133
Vibe Code Bench v1.1Coding
76.2% best41 / 123
70.4%48 / 123
Artificial Analysis SciCodeCoding
56.5%24 / 114
58.9% best13 / 114
Legal Research BenchLegal
48.1% best12 / 90
40.4%35 / 90
Tax Agent BenchFinance
33.7% best18 / 82
27.3%39 / 82
LiveBench Data AnalysisReasoning
75.3% best39 / 80
63.5%69 / 80
LiveBench ReasoningReasoning
91.2% best8 / 80
83.8%44 / 80
Finance Agent (v2)Finance
53.7% best32 / 79
50.7%42 / 79
Code MigrationCoding
44.6% best20 / 77
39.0%31 / 77
EMBFinance
62.7% best31 / 75
57.1%41 / 75
Terminal-Bench 4.0Agentic
17.2%37 / 61
31.8% best23 / 61
DeepSWEAgentic
65.9%33 / 59
67.7% best27 / 59
Agent Arena steerabilityAgentic
−1.834 / 54
−0.9 best30 / 54
Agent Arena task outcomeAgentic
−2.636 / 54
−1.5 best35 / 54
ProofBench v1.1Math
51.0% best32 / 54
42.0%37 / 54
LiveBench Coding 2.0Coding
89.1% best17 / 52
85.6%22 / 52