xAIPro TierZero Data Retention

Grok 4.5

Name: ARMES
Availability: InStock
Author: ARMES

Run Grok 4.5

Included on PRO plan

Context Window

500K

Input Token Price

$2.00

per 1,000,000 tokens

Output Token Price

$6.00

per 1,000,000 tokens

Speed Rating

Fast

Fast (~86.7 t/s output)

Verified Frontier Benchmarks

Empirical Evaluation Results

Evaluated across standard and verified harnesses

MoE ~1.5T params (co-trained with Cursor). Single SKU with configurable reasoning effort (low/medium/high; high default, non-disableable). Built-in server-side tools

science

web search

AA Intelligence Index

reasoning

38.8

AA Coding Index

coding

72.4

AnalystAgent

strategy

35.0%

AutomationBench

agentic

57.9%

AA Briefcase

strategy

1283.4 Elo

Critpt

reasoning

15.4%

EnterpriseOps Gym

agentic

40.8%

Gdp Pdf

strategy

18.8%

AA GDPval 2.0

strategy

1429.6 Elo

GPQA Diamond

reasoning

93.1%

Harvey LAB-AA

legal

92.4%

Humanity's Last Exam (HLE)

reasoning

42.7%

Lcr (1.1)

reasoning

79.3%

Mlcr

reasoning

15.0%

Mmmu Pro

reasoning

80.4%

Omniscience

science

25.3

SciCode

science

55.0%

Tau3 Banking

finance

42.1%

Terminal-Bench 2.1

coding

81.6%

Terminal-Bench 4.0

coding

10.6%

Coding Agent Index (1.4)

agentic

64.1%

Vals EMB 2.0

finance

52.9%

Vals Vibe Code Bench

coding

69.0%

Vals Code Migration

coding

38.4%

Vals Legal Research

legal

38.0%

FrontierCode 1.1

coding

42.4%

Bullshitbench V1

reasoning

60.0%

Bullshitbench V2

reasoning

55.0%

DeepSWE / SWE-bench Pro

coding

53.8%

Frontiermath Tiers 1 3 (v2)

math

57.2%

FrontierMath Tier 4

math

24.4%

SimpleQA

voice

48.3%

SWE Rebench

coding

63.8%

Chess Puzzles

reasoning

36.0%

Epoch GPQA Diamond

reasoning

93.4%

Posttrainbench (1.1)

reasoning

23.4%

Surge Chartography (100-tasks)

reasoning

17.0%

Surge Gdp Pdf (100-tasks)

strategy

14.0%

Vals Index (2)

reasoning

51.5%

Vals Finance Agent

finance

48.3%

Vals Terminal Bench 2.1 (2)

coding

67.8%

AA Agentic Index

agentic

42.1

TAU-Bench Banking

agentic

42.1%

Briefcase

strategy

1283.4 Elo

Gdpval (2)

strategy

1429.6 Elo

TAU-3 Banking

agentic

42.1%

DeepSWE v1.1

coding

53.8%

Design Arena Frontend

coding

1209 Elo

Design Arena Fullstack

coding

1247 Elo

Briefcase v1.1

strategy

1283.4 Elo

GDPval-AA v2.1

strategy

1429.6 Elo

AA-LCR v1.1

reasoning

79.3%

AA-Omniscience Index

reasoning

25.3

SimpleQA Verified

voice

48.3%

Terminal Bench (4.0)

coding

12.4%

Frontiercode Extended (1.1)

coding

56.5%

Furniture Assembly

finance

22.5%

Architectural Profile & Capabilities

Artificial Analysis Intelligence Index: 54 (#4 of 168), AA Coding Agent Index: 76 (on par with GPT-5.5), τ³-Banking: 33% (#1 of 28 models), GPQA Diamond: 93%, Terminal-Bench 2.1: 82–83.3%, SWE-Bench Pro: 64.7%, DeepSWE 1.0: 62%, SWE Marathon: 29% (beats Opus 4.8's 26%), ~15,954 output tokens per SWE-Bench Pro task (4.2× more token-efficient than Opus 4.8) | xAI's flagship model for software engineering, agentic execution, and technical reasoning (released Jul 8, 2026). **Strongest signals: production engineering and debugging, terminal/command-line workflows, end-to-end implementation, tool-driven agents, rapid prototyping, agentic banking/tool-use (#1), and high-throughput reasoning at lower cost.** 4.2× fewer tokens per SWE-Bench Pro task than Opus 4.8. 500K context, text + image input, text output. $2/$6 per MTok. **Caveats: hallucination rate 54% (vs 25% for 4.3); 500K context smaller than peers' 1M; evidence heavily engineering-focused. Do not auto-displace Opus for nuanced judgment, Sonnet for writing, or Sol for scientific reasoning.**

Architecture Type: Dense Transformer

Standard dense autoregressive transformer architecture with full-attention mechanisms.

Recommended Workloads & Primary Use Cases

Everyday reasoning

Drafting

Fast problem solving

Operational Caveats & Boundaries

•Extended context pricing applies above 200K tokens: $4/M in, $12/M out.

ARMES Zero Data Retention

Calls to Grok 4.5 are routed through strict Zero Data Retention inference channels. Prompts and outputs are never stored, indexed, or monitored by xAI.

View Privacy Architecture

More Models from xAI

Compare other engines available in this laboratory.

View Roster (5)

Grok 4.7Pro Tier

Grok 4.7 is SpaceXAI's flagship model for coding, agentic tasks, and knowledge work, succeeding Grok 4.6. Available with Zero Data Retention inference on ARMES AI.

500K ctx$2.00/M

Grok 4.6Pro Tier

AA Intelligence Index: 61 (matches GPT-5.6 Sol), GDPval-AA v2: 1753 Elo, CursorBench v3.2: 69.9%, DeepSWE v1.1: 65.9%, FrontierCode v1.1 Extended: 61.3%, APEX-Agents: 57.5%, Terminal-Bench v3.0: 26%, APEX-SWE: 56.4%, AA-Briefcase: 1577 Elo, Harvey LAB: 15.8%, SWE-bench Verified (Vals): 95.60%, GPQA Diamond: 94.9%, HLE: 42.9%, AA Coding Index: 76.8, AA Agentic Index: 58.7, AA Non-Hallucination Rate: 65.7% | xAI's current frontier model (released Aug 12, 2026; knowledge cutoff Feb 1, 2026).

500K ctx$2.00/M

Grok 4.3Pro Tier

CaseLaw v2: 79.3% (#1, +25 pts over 4.20), CorpFin: #1, τ²-Bench Telecom: 97.7%, GDPval-AA: 1500 ELO, GPQA Diamond: 90.1%, IFBench: 81.0%, SciCode: 47.3%, Coding Index: 41–42.2%, Terminal-Bench Hard: 38.0%, HLE: 35.0%, Artificial Analysis Intelligence Index: 53 | xAI's most capable reasoning model (beta Apr 17, 2026; GA Apr 30, 2026).

1M ctx$1.25/M

Architectural Profile & Capabilities

Architecture Type: Dense Transformer

Standard dense autoregressive transformer architecture with full-attention mechanisms.