VerdictPal
Join

Models / xAI

Grok 4.3

Grok · Released 2026-04-30

xAI's April 2026 frontier model with a 2M-token context window, real-time X search built in, and Vals Index 46.63%. Not the smartest model on the board, but the only one with a 2M context and a live social graph.

Default model for real-time social signals, very long context (>1M), and freshness-sensitive research where minutes matter.

#16 of 21 on Vals Index · current snapshot
47 Vals Index · 17 benchmark rows · 9 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite60
Context2M
Input / 1M$5.0
Output / 1M$25
Knowledge cutoff2026-03-15
Output speed124 t/s
TTFT (default API)31.32s
Statussolid
KnowledgeReasoningCodingAgenticLong contextTool useHuman preference
Knowledge
89
Reasoning
88
Coding
78
Agentic
64
Long context
64
Tool use
77
Human preference
88

78average across 7 tested families

Editor's note
Grok 4.3 is the only model on the public market with a real >1M context. The 2M window is real — but in our long-document tests, citation fidelity past 1.5M tokens drops sharply. The X search integration is a genuine differentiator for breaking-news research.
Benchmark placements

Where Grok 4.3 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
IFBenchReasoning#3/ 5781Artificial Analysis2026-07-21
LMArena (Chatbot Arena)Human preference#4/ 1388LMArena leaderboard2026-05-25
MMLUKnowledge#11/ 2289xAI Grok 4.32026-04-30
AA time to first tokenPerformance#12/ 69Artificial Analysis2026-07-21
BFCLTool use#12/ 2077Berkeley Function Calling Leaderboard V42026-04-15
Vals IndexAgentic#15/ 2047Vals AI — Vals Index2026-06-04
LiveBenchReasoning#17/ 2161LiveBench leaderboard2026-04-15
LongBench v2Long context#17/ 1946LongBench v2 leaderboard2026-04-01
SWE-bench VerifiedCoding#19/ 2378SWE-bench Verified leaderboard2026-06-09
AA output speedPerformance#21/ 69Artificial Analysis2026-07-21
Humanity's Last ExamReasoning#28/ 6535Artificial Analysis2026-07-21
GPQA DiamondReasoning#31/ 6988GPQA Diamond paper leaderboard2026-05-15
Terminal-BenchAgentic#33/ 6964Terminal-Bench 2.0 leaderboard2026-06-09
SciCodeCoding#34/ 6547Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#40/ 6738Artificial Analysis2026-07-21
AA-LCRLong context#41/ 6464Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#53/ 6412Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Grok 4.3 shows up, it's highlighted.

KnowledgeGrok 4.3 · 89
  1. 1
    GPT 5.5
    91
  2. 2
    Claude Opus 4.8
    91
  3. 3
    Claude Sonnet 4.6
    91
ReasoningGrok 4.3 · 88
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingGrok 4.3 · 78
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticGrok 4.3 · 64
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextGrok 4.3 · 64
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Tool useGrok 4.3 · 77
  1. 1
    Claude Sonnet 4.6
    89
  2. 2
    GPT 5.5
    87
  3. 3
    Claude Opus 4.8
    87
Human preferenceGrok 4.3 · 88
  1. 1
    GPT 5.5
    95
  2. 2
    Claude Opus 4.8
    92
  3. 3
    Gemini 3.5 Flash
    91
Newest receipts

Benchmark rows added to the public ledger for Grok 4.3 in the last 120 days. Older rows live in the full table below.

17 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context64.33%Artificial Analysis2
AA output speedPerformance124 t/sArtificial Analysis2
AA time to first tokenPerformance31.32sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning37.6%Artificial Analysis2
Humanity's Last ExamReasoning35%Artificial Analysis2
IFBenchReasoning81.29%Artificial Analysis2
SciCodeCoding47.3%Artificial Analysis2
τ³-Bench BankingAgentic12.16%Artificial Analysis2
SWE-bench VerifiedCoding77.6%SWE-bench Verified leaderboard44
Terminal-BenchAgentic63.8%Terminal-Bench 2.0 leaderboard44
Vals IndexAgentic46.63%Vals AI — Vals Index49
LMArena (Chatbot Arena)Human preference1441LMArena leaderboard59
GPQA DiamondReasoning88%GPQA Diamond paper leaderboard69
MMLUKnowledge88.5%xAI Grok 4.384
BFCLTool use77.2%Berkeley Function Calling Leaderboard V499
LiveBenchReasoning60.8%LiveBench leaderboard99
LongBench v2Long context46.2%LongBench v2 leaderboard113
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Knowledge89
Reasoning88
Coding78
Agentic64
Long context64
Tool use77
Human preference88
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Grok 4.3. Scores link to the original source; gaps mean no public row exists yet.

17 sourced rows

17 of 32 catalog benchmarks have a sourced row for Grok 4.3.

Knowledge

Reasoning

Coding

Agentic

Long context

Tool use

Performance

Human preference

Changelog
  1. Released Grok 4.3 with 2M context and real-time X search.
  2. Grok 4.20 retired from general availability.
Sources