Grok 4.5 Is Now Available on FastRouter

Grok 4.5 from xAI is now accessible through FastRouter. It is xAI's flagship multimodal reasoning model, designed for coding, agentic workflows, and knowledge work at aggressive per-token pricing for large, tool-heavy sessions.

What Grok 4.5 is

Grok 4.5 is positioned as an Opus-class model — meaning it targets the upper tier of capability alongside models like Claude Opus and GPT-5.6 Sol — but at pricing that is significantly more accessible for long, tool-heavy sessions.

The model accepts text and image inputs and produces text outputs. It supports tool calling for search, code execution, and other actions through the xAI API, making it well suited for agentic workflows where the model needs to take actions rather than just generate text.

How Grok 4.5 scores on independent benchmarks

Independent benchmarking from ArtificialAnalysis gives Grok 4.5 an Intelligence Index score of 54, ranking it #4 out of 170 tracked models. On coding specifically, it scores 72.4, placing it #6 on FastRouter's own model rankings. Both figures cover the reasoning-enabled variant of the model, since that is the only version ArtificialAnalysis currently benchmarks for Grok 4.5.

The pricing case

At $2.00 per million input tokens and $6.00 per million output tokens, Grok 4.5 is priced well below other Opus-class models. For agentic workflows that involve many tool calls and long context sessions, the per-token cost compounds quickly — lower input pricing at this level makes a meaningful difference in the economics of running production agents.

The context window of 500,000 tokens is large enough for most long-horizon tasks, including multi-document analysis, extended coding sessions, and agent loops that maintain substantial context across many steps.

What it is built for

Grok 4.5 is optimized specifically for three areas: coding, agentic workflows, and knowledge work. The tool calling support is a first-class feature — the model is designed to call tools naturally as part of its reasoning rather than treating tool calls as an afterthought.

The latency of under half a second to first token is notably fast for a model at this capability level, which matters for interactive agentic applications where the user is waiting on the model to decide what to do next. The tradeoff is throughput — at 133 tokens per second it generates text more slowly than some alternatives once it starts, which is worth factoring in for applications where output speed is the bottleneck.

Getting started

Grok 4.5 is available through the same FastRouter endpoint as every other model. No additional configuration is required.

1from openai import OpenAI
2
3client = OpenAI(
4    base_url="https://api.fastrouter.ai/api/v1",
5    api_key="<FASTROUTER_API_KEY>",
6)
7
8completion = client.chat.completions.create(
9    model="x-ai/grok-4.5",
10    messages=[
11        {"role": "user", "content": "Review this codebase and suggest refactors."}
12    ]
13)
14print(completion.choices[0].message.content)
Copy

Replace <FASTROUTER_API_KEY> with your actual FastRouter API key. Install the OpenAI SDK with pip install openai if not already installed.

Comparing Grok 4.5 against other models on FastRouter

The FastRouter Playground makes it straightforward to run the same prompt against Grok 4.5 and other models to compare output quality and cost side by side before committing to a routing decision in production.

Custom Evaluations can then be used to validate Grok 4.5 against a broader set of production prompts before switching any live traffic to it.

Grok 4.5 is available now through FastRouter at fastrouter.ai. One base URL, 160+ models, zero markup on API calls.