Back to the index
xAI · Grok
No. 17

Grok 4.1 Fast

Cheap, fast, 2M-context workhorse for high throughput.

FastClosed weightsText, Image
Field notes
Context window
2Mtokens
Max output
66Ktokens
Input price
$0.20/ 1M tokens
Output price
$0.50/ 1M tokens
Cached input
$0.05/ 1M tokens
Modalities
Text, Image
Knowledge cutoff
May 2025
Released
Nov 2025
Approx. speed
~250 tok/s
Overview

What this model is

Grok 4.1 Fast is xAI's high-throughput, low-cost tier with a 2M-token context window and aggressive prompt-cache pricing. It is great value for search, summarization, and high-volume calls with real-time data access. It trades away reliability on complex reasoning and multi-step coding, so it is best as a fast utility model rather than an autonomous coding agent.

Strengths

  • Very low cost with cheap prompt caching
  • Huge 2M-token context window
  • Fast response latency
  • Real-time X/web data access

Trade-offs

  • Weaker on complex coding and agentic tasks
  • Lower reasoning depth than the flagship
  • Less consistent tool use
  • Shallower on hard STEM problems

Best for

  • High-volume, low-cost inference
  • Retrieval and summarization
  • Real-time search over large context
  • Simple classification and extraction

Not ideal for

  • Complex multi-file coding
  • Long-horizon agentic automation
Capabilities

Capability profile

Normalized 0–100 scores, comparable across the whole catalog.

Reasoning74Coding68Math80Writing76Knowledge80Speed92Agentic66Vision74Multilingual76Long Context85
Reasoning
74
Coding
68
Math
80
Writing
76
Knowledge
80
Speed
92
Agentic
66
Vision
74
Multilingual
76
Long Context
85
Benchmarks

How it scores

Public benchmark results, with independent third-party results where available. Bars normalize percentages to 100 and Elo ratings to a 1500 ceiling.

Independent index
Artificial Analysis Intelligence Index
Composite of ~9–10 independent evals · Artificial Analysis
31/ 100
BenchmarkResult
Artificial Analysis Intelligence Indexindependent
GeneralArtificial Analysis
31%
τ-benchindependent
Agenticτ-bench
72.8%
Alternatives

Comparable models

Models in a roughly similar class — worth weighing against this one.