Grok 4.5
Frontier-Adjacent Coding, Priced to Run All Day
Grok 4.5 is xAI's flagship model for coding and agentic work, which xAI calls the most intelligent and fastest model it has built. It reads 500,000 tokens of context and reaches its answers using a fraction of the output tokens rival models spend. Available on Pexo.
See What You Can Make with Pexo
Here's a glimpse of what people create with Pexo — just a description, no setup.
What Makes Grok 4.5 Different
xAI shipped Grok 4.5 on July 8, 2026. What makes it interesting is not a leaderboard position, it is how it was trained and what it costs to run.
Taught Inside a Real Editor, Not on Isolated Snippets
Grok 4.5 is the first model xAI and Cursor trained together, using real developer session data from Cursor and evaluations designed to measure what a model can actually do inside a live codebase over a long session. Training ran across tens of thousands of NVIDIA GB300 GPUs with reinforcement learning over hundreds of thousands of multi-step software engineering tasks.
It Gets There Without Thinking Out Loud Forever
This is the number xAI leads with. On SWE-bench Pro it reports Grok 4.5 resolving tasks in roughly 15,954 output tokens on average, against about 67,020 for Claude Opus 4.8 on the same benchmark. Since output tokens are what you actually pay for on long agentic runs, that gap compounds across a working day.
500K of Context at $2 per Million In
Grok 4.5 reads a 500,000 token context window and lists at $2 per million input tokens and $6 per million output, with cached input at $0.30. One thing to watch: once a request crosses 200,000 prompt tokens, the whole request bills at the long-context rate of $4 and $12, so batching matters.
Grok 4.5 vs the Models It Is Priced Against
Elon Musk pitched Grok 4.5 as Opus-class but faster, more token-efficient and lower cost, and his more careful phrasing put it at roughly comparable to Opus 4.7. Here is where it actually sits.
| Capability | Grok 4.5 | Claude Opus 5 | GPT-5.6 Sol | Kimi K3 |
|---|---|---|---|---|
| Listed input price per million | $2 | $5 | Not disclosed | $3 |
| Context window | 500K | Not disclosed | Not disclosed | 1M |
| Coding focus in positioning | Primary | Primary | Primary | Primary |
| Output token efficiency | Frontier | Strong | Frontier | Standard |
| Weights published | — | — | — | ✓ |
| Reachable through Pexo | ✓ | ✓ | ✓ | ✓ |
Sources: xAI: Models and pricing · xAI: Grok 4.5 · xAI
How to Use Grok 4.5 with Pexo
Grok 4.5 is built to run inside an agent harness. Give it the Pexo skill and it can make video for you, not just talk about it.
Run Grok 4.5 where it already works — Claude Code, Codex, or OpenClaw — and install the Pexo skill. Pexo exposes generation as a skill plus an API, so the agent gains video as a capability it can call.
Stay in the same conversation you're already working in. Describe the shot, the mood, or the whole story in plain language. No prompt syntax and no separate tool to open.
Grok 4.5 turns your intent into the right call, Pexo generates the video, and the file comes back into your workflow. Want a different take? Say so and it iterates.
What You Can Make with Pexo
No prompt writing. No model picking. Just describe what you want.
Frequently Asked Questions About Grok 4.5
What is Grok 4.5?+
Grok 4.5 is xAI's flagship language model, released on July 8, 2026 and built for coding, agentic tasks, and knowledge work. xAI's own documentation calls it the most intelligent and fastest model it has built and recommends it for code and general chat. In the API it is called grok-4.5.
How much does Grok 4.5 cost?+
$2 per million input tokens and $6 per million output, with cached input at $0.30. Requests whose prompt reaches 200,000 tokens bill at the long-context tier instead, $4 input and $12 output, and that rate applies to the whole request rather than just the tokens past the threshold.
What does the Cursor partnership actually mean?+
Grok 4.5 is the first model xAI and Cursor trained jointly. Cursor contributed real developer session data, and the evaluations were designed around long sessions inside real repositories rather than isolated coding puzzles. That is why its strongest claims are about sustained engineering work rather than single-shot benchmarks.
Is Grok 4.5 better than Claude or GPT?+
Not on raw capability. Independent scoring from Artificial Analysis placed it fourth at launch, behind Fable 5, GPT-5.5, and Opus 4.8. Its real argument is intelligence per dollar: comparable work at a fraction of the price and a fraction of the output tokens. Treat xAI's token efficiency figures as its own published numbers until someone re-runs them.
How big is the Grok 4.5 context window?+
500,000 tokens, with a knowledge cut-off of February 1, 2026. Grok 4.5 itself handles text and code. Image, video, and voice tasks go to xAI's separate Grok Imagine and Grok Voice APIs rather than being native to this model.
How do I get access to Grok 4.5?+
Through the xAI API as grok-4.5, and in the Grok app for SuperGrok and X Premium+ subscribers. Availability rolled out later in the EU than elsewhere, so check your region. If you want it to make video as well as reason about it, add the Pexo skill to the agent you already run it in, such as Claude Code, Codex, or OpenClaw, and Pexo becomes a capability the model can call.
Give Grok 4.5 Video with Pexo
Add the Pexo skill to Claude Code, Codex, or OpenClaw, then describe the video you want and let the model make it.