Technical Specifications of Grok 4.5

Item	Specification
Model Name	Grok 4.5
Provider	xAI
API Model ID	grok-4.5
Model Type	Large Language Model (LLM)
Input Modalities	Text, Images
Output Modalities	Text
Primary Strengths	Coding, agentic workflows, software engineering, enterprise productivity, knowledge work
API Access	xAI API, Grok Build, Cursor integration, CometAPI
Reasoning	Multi-step reasoning optimized for engineering tasks
Release Date	July 2026

Specifications are based on xAI's official launch announcement and developer documentation.

What Is Grok 4.5?

Grok 4.5 is xAI's most advanced foundation model, purpose-built for software engineering, autonomous agent workflows, enterprise productivity, and complex technical reasoning. Unlike earlier Grok releases that emphasized conversational AI, Grok 4.5 focuses heavily on real-world engineering performance, including coding, debugging, document generation, spreadsheet automation, and multi-step task execution.

The model was trained alongside Cursor using large-scale engineering datasets covering software development, mathematics, science, and technical documentation. xAI also introduced significantly improved reinforcement learning and data curation pipelines to improve reasoning efficiency while reducing unnecessary token generation.

Features and Highlights of the Grok 4.5 API

1. Built for Professional Software Engineering

Grok 4.5 is optimized for practical coding rather than benchmark-only performance. It can:

Generate production-ready applications
Debug large codebases
Refactor existing software
Explain complex algorithms
Write Rust, C/C++, Python, JavaScript, Go, Java, and more

According to xAI, many complete applications showcased during launch were generated from a single prompt.

2. Excellent Agentic Workflows

Grok 4.5 is designed for long-running autonomous tasks where the model plans, reasons, and iteratively improves results.

Typical agent tasks include:

Repository migrations
Automated bug fixing
Documentation generation
Multi-file code editing
Research and synthesis
Workflow automation

Its reinforcement learning pipeline specifically targets long-duration engineering tasks rather than short benchmark questions.

3. Enterprise Office Productivity

Beyond coding, Grok 4.5 powers Grok Build for creating:

Complex Excel workbooks
PowerPoint presentations
Word documents
Business reports
Financial models
Multi-sheet spreadsheet analysis

The model can combine web research with document creation, making it useful for business analysts, consultants, finance teams, and operations professionals.

4. Faster Reasoning with Lower Token Usage

A major design goal of Grok 4.5 is efficiency.

According to xAI:

Up to 80 tokens per second generation speed
Roughly 4.2× fewer output tokens than comparable leading models on SWE Bench Pro
Lower inference costs for long engineering workflows

Instead of maximizing reasoning length, Grok 4.5 focuses on producing concise solutions while maintaining high task accuracy.

Benchmark Performance

xAI positions Grok 4.5 among the strongest engineering-focused models available.

Benchmark	Grok 4.5
DeepSWE 1.0	62.0%
Terminal Bench 2.1	83.3%
SWE Bench Pro	64.7%
Harvey Legal Agent Benchmark	#1 (reported by xAI)

While Grok 4.5 trails some competitors on selected coding benchmarks, it combines competitive engineering performance with significantly higher inference speed and token efficiency.

Preise für Grok 4.5

Entdecken Sie wettbewerbsfähige Preise für Grok 4.5, die für verschiedene Budgets und Nutzungsanforderungen konzipiert sind. Unsere flexiblen Tarife stellen sicher, dass Sie nur für das bezahlen, was Sie nutzen, und erleichtern die Skalierung entsprechend Ihren wachsenden Anforderungen. Erfahren Sie, wie Grok 4.5 Ihre Projekte verbessern kann, während die Kosten überschaubar bleiben.

Comet-Preis (USD / M Tokens)	Offizieller Preis (USD / M Tokens)	Rabatt
Eingabe:$1.6/M Ausgabe:$3.2/M	Eingabe:$2/M Ausgabe:$4/M	-20%

Beispielcode und API für Grok 4.5

Greifen Sie auf umfassende Beispielcodes und API-Ressourcen für Grok 4.5 zu, um Ihren Integrationsprozess zu optimieren. Unsere detaillierte Dokumentation bietet schrittweise Anleitungen und hilft Ihnen dabei, das volle Potenzial von Grok 4.5 in Ihren Projekten zu nutzen.

Python
JavaScript
Curl

import os

from openai import OpenAI

# Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"
BASE_URL = "https://api.cometapi.com/v1"

client = OpenAI(base_url=BASE_URL, api_key=COMETAPI_KEY)
response = client.responses.create(
    model="grok-4.5",
    input=[
        {
            "role": "user",
            "content": "Find and fix the bug, then explain it: function median(a){a.sort();return a[a.length/2]}",
        }
    ],
)

print(response.output_text or response.model_dump_json(indent=2))

Python Code Example

import os

from openai import OpenAI

# Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"
BASE_URL = "https://api.cometapi.com/v1"

client = OpenAI(base_url=BASE_URL, api_key=COMETAPI_KEY)
response = client.responses.create(
    model="grok-4.5",
    input=[
        {
            "role": "user",
            "content": "Find and fix the bug, then explain it: function median(a){a.sort();return a[a.length/2]}",
        }
    ],
)

print(response.output_text or response.model_dump_json(indent=2))

JavaScript Code Example

import OpenAI from "openai";

// Get your CometAPI key from https://www.cometapi.com/console/token, and paste it here
const apiKey = process.env.COMETAPI_KEY || "<YOUR_COMETAPI_KEY>";
const baseUrl = "https://api.cometapi.com/v1";

const client = new OpenAI({
  apiKey,
  baseURL: baseUrl,
});

const response = await client.responses.create({
  model: "grok-4.5",
  input: [
    {
      role: "user",
      content: "Find and fix the bug, then explain it: function median(a){a.sort();return a[a.length/2]}",
    },
  ],
});

console.log(response.output_text ?? JSON.stringify(response.output, null, 2));

Curl Code Example

#!/usr/bin/env bash
# Get your CometAPI key from https://www.cometapi.com/console/token
# Export it as: export COMETAPI_KEY="your-key-here"

response=$(curl --silent --location --request POST "https://api.cometapi.com/v1/responses" \
  --header "Authorization: Bearer $COMETAPI_KEY" \
  --header "Content-Type: application/json" \
  --header "Accept: application/json" \
  --data-raw '{
    "model": "grok-4.5",
    "input": [
      {
        "role": "user",
        "content": "Find and fix the bug, then explain it: function median(a){a.sort();return a[a.length/2]}"
      }
    ]
  }')

if command -v jq >/dev/null 2>&1; then
  printf '%s\n' "$response" | jq -r '(
    [
      .output[]?
      | select(.type == "message")
      | .content[]?
      | select(.type == "output_text")
      | .text
    ][0]
  ) // .output_text // .'
else
  printf '%s\n' "$response"
fi

Versionen von Grok 4.5

Der Grund, warum Grok 4.5 mehrere Snapshots hat, kann potenzielle Faktoren wie Änderungen der Ausgabe nach Updates umfassen, die ältere Snapshots für Konsistenz erfordern, Entwicklern eine Übergangszeit für Anpassung und Migration bieten und verschiedene Snapshots, die globalen oder regionalen Endpunkten entsprechen, um das Benutzererlebnis zu optimieren. Für detaillierte Unterschiede zwischen den Versionen lesen Sie bitte die offizielle Dokumentation.

version
grok-4.5