Patchwire

AI lab · 7 models tracked

xAI models

Every xAI model released in the last 18 months, with release dates, context windows, API prices and benchmark scores. Newest: Grok 4.7, 29d ago.

Compare xAI models → All labs
Models · 30 days
1 of 7
Latest release
29d ago
Cheapest input / 1M
$1.00 Grok Build 0.1
Largest context
2M Grok 4.20 Multi-Agent

Release history

Timeline

xAI ships a new model about every 29 days. Each dot is a release; click one to open it.

Jul 2025Oct 2025Jan 2026Apr 2026Jul 2026 Grok 4.20 — Mar 31, 2026Grok 4.20 Multi-Agent — Mar 31, 2026Grok 4.3 — Apr 30, 2026Grok Build 0.1 — May 20, 2026Grok 4.5 — Jul 8, 2026Grok 4.6 — Aug 12, 2026Grok 4.7 — Sep 21, 2026

Lineup

All xAI models

ModelReleasedContextInput / 1MOutput / 1MGPQA Diamond
Grok 4.7x-ai/grok-4.7
ReasoningVision
 Sep 21, 2026 500K450K out $2.00 $6.00 92.7%
Grok 4.6x-ai/grok-4.6
ReasoningVision
 Aug 12, 2026 500K450K out $2.00 $6.00 94.0%
Grok 4.5x-ai/grok-4.5
ReasoningVision
 Jul 8, 2026 500K450K out $2.00 $6.00 93.4%
Grok Build 0.1x-ai/grok-build-0.1
ReasoningVision
 May 20, 2026 256K230K out $1.00 $2.00 —
Grok 4.3x-ai/grok-4.3
ReasoningVision
 Apr 30, 2026 1M900K out $1.25 $2.50 88.8%
Grok 4.20 Multi-Agentx-ai/grok-4.20-multi-agent
ReasoningVision
 Mar 31, 2026 2M1.8M out $1.25 $2.50 —
Grok 4.20x-ai/grok-4.20
ReasoningVision
 Mar 31, 2026 2M1.8M out $1.25 $2.50 —

Benchmarks

xAI's best scores

The top xAI model on each benchmark, and where it ranks among all 222 tracked models.

BenchmarkBest xAI modelScoreOverall rank
FrontierSWESoftware engineering Grok 4.7 29.5% #11of 19
DeepSWESoftware engineering Grok 4.6 67.5% #9of 26
CursorBenchSoftware engineering Grok 4.7 46.3% #7of 16
FrontierCodeSoftware engineering Grok 4.6 48.0% #9of 41
Terminal-BenchSoftware engineering Grok 4.20 57.3% #14of 40
GBAEvalSoftware engineering Grok 4.5 65.4% #4of 23
SciCodeSoftware engineering Grok 4.7 57.8% #14of 93
WeirdML v3Software engineering Grok 4.7 8.5% #12of 22
APEX-AgentsAgents & computer use Grok 4.6 65.3% #7of 51
Vending-Bench 2Agents & computer use Grok 4.7 $10,537 #7of 61
PostTrainBenchAgents & computer use Grok 4.5 23.4% #9of 11
GDP.pdfWriting & content Grok 4.7 22.8% #19of 38
SimpleQA VerifiedWriting & content Grok 4.7 56.0% #19of 66
WebDev ArenaDesign & visual Grok 4.7 1638 #12of 88
Blueprint-Bench 2Design & visual Grok 4.6 33.2% #12of 32
Furniture AssemblyDesign & visual Grok 4.6 40.0% #16of 34
GPQA DiamondReasoning & knowledge Grok 4.6 94.0% #11of 92
ARC-AGI-2Reasoning & knowledge Grok 4.6 67.1% #24of 63
ARC-AGI-1Reasoning & knowledge Grok 4.7 90.2% #30of 63
SimpleBenchReasoning & knowledge Grok 4.6 75.9% #6of 43
CritPtReasoning & knowledge Grok 4.6 19.7% #25of 96
Chess PuzzlesReasoning & knowledge Grok 4.7 41.0% #35of 70
LMCAReasoning & knowledge Grok 4.7 49.4% #27of 96
FrontierMathMaths Grok 4.6 66.0% #31of 72
FrontierMath Tier 4Maths Grok 4.6 31.7% #26of 62
ProofBenchMaths Grok 4.6 51.0% #30of 70
OTIS Mock AIMEMaths Grok 4.6 99.2% #12of 90

Directory

Other labs