Grok 4.5

Name: Grok 4.5
Author: xAI

xAI · 2026-07-08

Grok 4.5 is SpaceXAI's proprietary flagship model for coding, agentic tasks, and knowledge work, released July 8, 2026. It was trained alongside Cursor on curated coding, science, engineering, and mathematics data, with reinforcement learning across hundreds of thousands of multi-step software-engineering and technical tasks. The model supports text and image input, a 500,000-token context window, and configurable reasoning.

Benchmark scores

Benchmark	Score
AA-AnalystAgent (pass@1)	54.0
AA-AnalystAgent (pass@5)	76.3
AA-AnalystAgent (pass^5)	35.0
AA-Briefcase Elo	1283
AA-Briefcase v1.1 Elo	1278
AA-LCR	67.7
AA-LCR v1.1	79.3
AA-Omniscience Index	25.3
AdvancedIF	79.3
AdvancedIF (rubric mean score)	95.0
Agent Red Teaming k=100 (CAIS)	83.5
AHK-Eval	86.1
AHK-Eval Algorithms	66.7
AHK-Eval Data Structures	83.3
AHK-Eval Date and Time	83.3
AHK-Eval Easy	100.0
AHK-Eval Hard	75.0
AHK-Eval Hidden Cases	88.4
AHK-Eval Mid	83.3
AHK-Eval Numbers	100.0
AHK-Eval Parse Success	94.4
AHK-Eval Regex	100.0
AHK-Eval Strings	100.0
Aikido Known-CVE Detection (Pass@3 Union)	76.9
Aikido Known-CVE Detection (Single-Run Average)	66.7
AIME 2025	98.3
ALE-Bench	1467
Antidote: Everyday Edition	973
APEX (Mercor)	11.9
APEX-Accounting	43.7