이 모델의 강점은 무엇인가요?

Advanced reasoning capabilities Broad 200K context window Cutting-edge design by xAI

이 모델의 약점은 무엇인가요?

Non-open source license Lack of detailed performance metrics Limited accessibility/environments

어떤 용도에 가장 적합한가요?

Complex logical reasoning Long-form document analysis Advanced problem solving

모델 목록으로

xAI독점

Grok 4.2

Name: Grok 4.2
Author: xAI

Grok 4.2 is a reasoning model developed by xAI. It features a long 200K context window and provides advanced reasoning capabilities.

파라미터

Undisclosed

컨텍스트

200K

라이선스

Proprietary

출시일

2026-02-17

API 가격

이 모델의 API 가격 정보는 현재 공개되지 않았습니다

강점

・Advanced reasoning capabilities
・Broad 200K context window
・Cutting-edge design by xAI

약점

・Non-open source license
・Lack of detailed performance metrics
・Limited accessibility/environments

활용 사례

・Complex logical reasoning
・Long-form document analysis
・Advanced problem solving

심층 분석

Arena Elo

~1493

#4 overall (preliminary)

Output Speed

234.9 tok/s

#1 among flagships

Context Window

2M tokens

largest of any flagship

Input Price

$2/1M

cheapest flagship

Output Price

$6/1M

cheapest flagship

Multi-Agent

16 sub-agents

native multi-agent mode

강점

・Largest context window (2M tokens) among flagship models
・Fastest output speed (234.9 tok/s) — 3x faster than Claude
・Cheapest flagship pricing ($2/$6 per 1M tokens)
・Native multi-agent mode with up to 16 coordinating sub-agents

약점

・No published SWE-bench or GPQA scores — limited benchmark transparency
・Multi-agent runs can be expensive (all tokens billed)
・Trails Claude Opus 4.6 on Arena Elo (~1493 vs ~1500)

경쟁사 비교

Model	Arena	Price
Claude Opus 4.6	~1500	$15/$75
Grok 4.2	~1493	$2/$6
GPT-5.4	1467	$2.50/$15
Gemini 3.1 Pro	~1485	$2/$10

개요

Grok 4.2 (also known as Grok 4.20) is xAI's flagship model, launched in public beta on February 17, 2026. It represents a major upgrade over Grok 4, with an 8x larger context window (2M tokens), a native multi-agent mode with up to 16 coordinating sub-agents, and significantly lower pricing ($2/$6 vs $3/$15 for Grok 4). The model's standout feature is its output speed — 234.9 tokens per second, roughly 3x faster than Claude Opus 4.6. Combined with the 2M context window, this makes Grok 4.2 uniquely suited for long-document processing and high-throughput workloads. At $2/$6 per 1M tokens, Grok 4.2 is the cheapest flagship model available. The Arena Elo of ~1493 (#4) puts it competitively close to Claude Opus 4.6 (~1500) and ahead of GPT-5.4 (1467).

벤치마크 및 성능

## Benchmark Scores | Benchmark | Grok 4.2 | Claude Opus 4.6 | GPT-5.4 | Gemini 3.1 Pro | |-----------|----------|-----------------|---------|----------------| | Arena Elo | ~1493 | ~1500 | 1467 | ~1485 | | Output Speed | 234.9 tok/s | ~80 tok/s | ~100 tok/s | ~120 tok/s | | Context Window | 2M | 1M | 1.05M | 1M | | Input Price | $2/1M | $15/1M | $2.50/1M | $2/1M | | Output Price | $6/1M | $75/1M | $15/1M | $10/1M | Note: xAI has not published SWE-bench, GPQA, or other standard academic benchmarks for Grok 4.2. The Arena Elo is preliminary based on ~5,071 votes.

상세 비교

## Grok 4.2 vs Claude Opus 4.6 Claude Opus 4.6 leads on Arena Elo (~1500 vs ~1493) and has published benchmark scores. Grok 4.2 counters with 2x larger context (2M vs 1M), 3x faster output (234 vs 80 tok/s), and dramatically lower pricing ($2/$6 vs $15/$75). ## Grok 4.2 vs GPT-5.4 Grok 4.2 offers a larger context window (2M vs 1M), faster output (234 vs 100 tok/s), and lower pricing ($2/$6 vs $2.50/$15). GPT-5.4 has more published benchmarks and proven agentic coding performance. ## Grok 4.2 vs Gemini 3.1 Pro Similar pricing ($2/$6 vs $2/$10) and context window (2M vs 1M). Grok 4.2 is faster; Gemini 3.1 Pro has more benchmark data.

커뮤니티 평가

Grok 4.2 received strong interest for its 2M context window and aggressive pricing. The native multi-agent mode was seen as innovative, though developers noted that all agent tokens are billed, making complex runs potentially expensive. The output speed of 234.9 tok/s was universally praised — it's genuinely 3x faster than any competitor at the flagship tier. For high-throughput workloads, this translates directly to lower latency and higher throughput. The lack of published standard benchmarks (SWE-bench, GPQA) was a concern for some teams, who prefer models with transparent, reproducible evaluations.

활용 사례

1. **Long-Document Processing**: 2M context window enables loading entire codebases, multi-document legal review, or extended conversation histories without retrieval. 2. **High-Throughput Workloads**: 234.9 tok/s output speed makes it ideal for real-time applications, streaming responses, and high-volume API workloads. 3. **Multi-Agent Research**: Native 16-agent mode enables complex research tasks where multiple perspectives are explored simultaneously. 4. **Cost-Sensitive Production**: At $2/$6, it's the cheapest flagship model, making it suitable for high-volume production deployments.