xAIPro TierZero Data Retention

Grok 4.20

Name: ARMES
Availability: InStock
Author: ARMES

Run Grok 4.20

Included on PRO plan

Context Window

2M (API) / 1M (playground); 30K max output (API) / 131K (playground)

Input Token Price

$1.25

per 1,000,000 tokens

Output Token Price

$2.50

per 1,000,000 tokens

Speed Rating

Fast

Fast (reasoning + non-reasoning variants)

Verified Frontier Benchmarks

Empirical Evaluation Results

Evaluated across standard and verified harnesses

MoE ~3T params on Colossus supercluster. "Hydra Backbone" native multi-agent

agentic

Grok (coordinator)

Long Horizon Terminal Bench (1.0)

coding

0.1

Architectural Profile & Capabilities

SWE-bench Verified: 78.0%, GPQA Diamond: 82.7%, MMLU-Pro: 83.7%, AIME 2025: 90.2%, IFBench: 72.6%, SciCode: 42.0%, Terminal-Bench Hard: 31.0%, τ²-Bench Telecom: 64.4%, HLE: 22.8–30.0%, non-hallucination: ~78% (Thinking mode) | xAI's large-context multi-agent workhorse (beta Feb 17, 2026; GA Mar 10, 2026). **Best for: real-time financial sentiment & trend parsing (direct 𝕏 firehose, ~68M posts/day), large-repository/multi-document ingestion in a single 2M-token prompt, reduced hallucinations via adversarial debate, and long narrative continuity.** Two configs: `grok-4.20-...-reasoning` (deliberative) and `-non-reasoning` (low-latency). Output pricing ~60% cheaper than GPT-5.4. **Caveats for ARMES: high ODCV-Bench misalignment (62.8%) under KPI pressure — use external state-aware validators before executing model-generated commands/file edits; prone to "brittle safety" refusals, defensive/adversarial loops in long sessions, occasional agent-sync failures, and content/political bias from heavy 𝕏 reliance. Not recommended for unsupervised autonomous agents.**

Architecture Type: Dense Transformer

Standard dense autoregressive transformer architecture with full-attention mechanisms.

Recommended Workloads & Primary Use Cases

Everyday reasoning

Drafting

Fast problem solving

ARMES Zero Data Retention

Calls to Grok 4.20 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 4.5Pro Tier

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).

500K ctx$2.00/M

Open App

Back to xAI models roster

xAIPro TierZero Data Retention

Grok 4.20

Run Grok 4.20

Included on PRO plan

Context Window

2M (API) / 1M (playground); 30K max output (API) / 131K (playground)

Input Token Price

$1.25

per 1,000,000 tokens

Output Token Price

$2.50

per 1,000,000 tokens

Speed Rating

Fast

Fast (reasoning + non-reasoning variants)

Verified Frontier Benchmarks

Empirical Evaluation Results

Evaluated across standard and verified harnesses

MoE ~3T params on Colossus supercluster. "Hydra Backbone" native multi-agent

agentic

Grok (coordinator)

Long Horizon Terminal Bench (1.0)

coding

0.1

Architectural Profile & Capabilities

Architecture Type: Dense Transformer

Standard dense autoregressive transformer architecture with full-attention mechanisms.

Recommended Workloads & Primary Use Cases

Everyday reasoning

Drafting

Fast problem solving

ARMES Zero Data Retention

Calls to Grok 4.20 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.