Atlas

Models

← All models

Grok 4.3

xAI · 2026-04-30

Grok 4.3 is xAI's flagship model released in April 2026, with a 1 million-token context window and configurable reasoning. xAI's current docs describe it as their most advanced flagship model, focused on low hallucination rates, agentic tool calling, and instruction following, with text and image inputs and text output.

Benchmark scores

BenchmarkScore
AA-AnalystAgent (pass@1)31.8
AA-AnalystAgent (pass@5)57.5
AA-AnalystAgent (pass^5)8.8
AA-Briefcase Elo762
AA-Briefcase v1.1 Elo763
AA-LCR65.0
AA-LCR v1.175.0
AA-Omniscience Index18.0
Agent Red Teaming k=100 (CAIS)90.1
Agents' Last Exam (ALE)7.2
Agents' Last Exam ALE-CLI Pass Rate8.6
Agents' Last Exam ALE-CLI Score25.9
Agents' Last Exam Full-Spectrum Pass Rate7.3
Agents' Last Exam Full-Spectrum Score18.3
Agents' Last Exam Last-Exam Pass Rate0.0
Agents' Last Exam Last-Exam Score2.3
Agents' Last Exam Near-Term Pass Rate10.4
Agents' Last Exam Near-Term Score31.8
Agents' Last Exam Overall Score21.3
Agents' Last Exam Unlicensed Full-Spectrum Pass Rate8.0
Agents' Last Exam Unlicensed Full-Spectrum Score19.9
Agents' Last Exam Unlicensed Last-Exam Pass Rate0.0
Agents' Last Exam Unlicensed Last-Exam Score3.0
Agents' Last Exam Unlicensed Near-Term Pass Rate10.4
Agents' Last Exam Unlicensed Near-Term Score31.8
Agents' Last Exam Unlicensed Overall Pass Rate7.8
Agents' Last Exam Unlicensed Overall Score22.8
AHK-Eval86.1
AHK-Eval Algorithms83.3
AHK-Eval Data Structures50.0
Loading Atlas data…