🎧 Prefer to listen?
The richest men in tech are fighting again — and this time, I decided to actually test what they’re selling.
Sam Altman and Elon Musk have been throwing punches at each other all week. Literally. On July 11, Altman boasted that OpenAI’s latest model is “the best in the world,” and Musk fired back accusing him of “scamming to a whole new level.” Meanwhile, Apple is suing OpenAI over alleged trade secrets, and Musk is siding with Apple. It’s messy. It’s personal. And it’s completely irrelevant to whether the tools actually work.
So I ignored the drama and spent two hours with Grok 4.5 — the model Musk calls “Opus-class, much faster.” There’s a 50% off promo running through July 15, which made it easy to justify testing. Here’s what I found.
What Even Is Grok Now?
If you haven’t followed the xAI saga — you’re not alone. The company that built Grok was folded into SpaceX in May 2026, rebranded as SpaceXAI on July 6, and launched Grok 4.5 two days later. The product still calls itself Grok. The pricing page still lives at x.ai. But the entity behind it is now SpaceX’s AI division.
The pricing tiers are genuinely confusing — there are eight of them, ranging from free to $300/month. Here’s the quick version:
- Free — text-only, limited prompts, no image generation
- X Premium ($8/mo) — light Grok use bundled with your X subscription
- SuperGrok Lite ($10/mo) — cheapest standalone entry
- SuperGrok ($30/mo) — full model access, image generation, 128K context
- SuperGrok Heavy ($300/mo) — multi-agent mode, maximum compute
If you’re comparing this to ChatGPT alternatives, the pricing is actually competitive. SuperGrok at $30 matches ChatGPT Plus, and the free tier is more generous than most — about 10 prompts every two hours.
The 50% Off Promo — Is It Real?
Yes. Through July 15, SuperGrok is available at half price. That’s $15/month instead of $30. If you’ve been curious about Grok, this is the cheapest way to test it with full features. The promo doesn’t apply to SuperGrok Heavy — you’re still paying $300 for that — but for everyday use, $15 is hard to argue with.
I signed up, plugged in my usual tasks, and started comparing.
What Grok Actually Does Well
Speed. This is the most noticeable thing. Grok 4.5 is fast. Not “fast for an AI” — actually fast. Responses come back in under five seconds for most tasks, including reasoning-heavy ones. If you’ve been frustrated by Claude’s processing delays or ChatGPT’s occasional sluggishness during peak hours, Grok feels snappier.
Understanding messy input. This is where the Twitter training data shows up. I tested it with half-finished thoughts, voice-dictated prompts, and vague descriptions of what I wanted. Grok understood intent faster than any model I’ve used. It reads between the lines in a way that feels different from Claude’s more literal interpretation or ChatGPT’s sometimes overly cautious clarifying questions.
If you’ve read my breakdown of why your AI output sucks, you know that input quality matters more than model choice. But Grok is more forgiving of messy input than most. It pattern-completes what you meant, not just what you said.
Price-to-capability ratio. At $2 per million input tokens and $6 per million output tokens on the API, Grok 4.5 is roughly half the price of Claude Opus 4.8 and GPT-5.6 Sol. Independent testing from Artificial Analysis ranks it fourth on their intelligence index — not the smartest, but the best value at the frontier tier.
Image generation. Grok’s image generation is solid. Not Midjourney-level for artistic work, but for quick visuals, social media graphics, and concept mockups, it’s fast and the results are usable. I tested it against the free AI image generators I’ve covered before, and Grok holds its own.
Where It Falls Short
Hallucination rate. This is the real caveat. Independent measurements flag Grok’s hallucination rate as high compared to Claude and ChatGPT. For factual lookups, research, or anything where accuracy matters — double-check the output. It’s not unreliable, but it’s confident when it shouldn’t be.
Reasoning depth. Grok 4.5 ranks fourth on the Artificial Analysis Intelligence Index, behind Claude Fable 5, GPT-5.5, and Opus 4.8. For complex multi-step reasoning — the kind of tasks where you’re asking an AI to work through a problem with five variables — Claude and GPT still edge it out. For everyday tasks, you won’t notice the difference. For building automations or complex workflows, you might.
EU availability. If you’re in Europe, Grok is blocked. That’s a dealbreaker for a chunk of my readers, and it’s worth knowing before you get invested.
Ecosystem maturity. ChatGPT has plugins, Claude has projects, Gemini has Google integration. Grok has… X integration. If you live on X/Twitter, that’s useful. If you don’t, it’s not much of an ecosystem advantage. The tool integrations are still catching up.
The Model Quality Drop — Am I Imagining This?
One thing I noticed, and this matches a pattern I’ve seen with other launches: Grok 4.5 feels better right now than it probably will in six weeks.
This isn’t a conspiracy theory. It’s economics. Companies launch models at full compute to make a splash. Once adoption hits targets, they optimize for throughput — more users, less compute per request, slightly worse output. The model name stays the same. The quality quietly drops. I watched the same thing happen with Composer 2.5 — incredible at launch, noticeably worse two months later.
I’m not saying Grok will definitely degrade. I’m saying if you’re going to test it, test it now while the promo is running and the model is fresh. Save your prompts. Compare in a month. That’ll tell you if it’s real.
Grok vs. ChatGPT vs. Claude — The Honest Comparison
I covered this in depth in my ChatGPT alternatives guide, but here’s the quick version after testing Grok specifically:
| Grok 4.5 | ChatGPT (GPT-5.6) | Claude (Opus 4.8) | |
|---|---|---|---|
| Speed | Fastest | Variable | Slower |
| Messy input handling | Best | Good | Literal |
| Reasoning | Good | Great | Best |
| Hallucination | Higher | Moderate | Lowest |
| Price | Cheapest | Mid | Most expensive |
| Ecosystem | X-focused | Broadest | Growing |
If you want the fastest, most forgiving model for everyday tasks — Grok is genuinely good. If you need precision and depth — Claude is still king. If you want the broadest ecosystem — ChatGPT wins.
Who Should Actually Switch?
Try Grok if:
- You’re on a budget and want frontier-tier AI at half price
- You dictate prompts via voice and want a model that handles messy input
- Speed matters more than peak reasoning depth
- You already use X and want integrated AI
Stick with Claude or ChatGPT if:
- You need the lowest hallucination rate for professional work
- You’re in the EU
- You rely on deep integrations with other tools
- You need multi-step reasoning for complex projects
For most non-technical users — the people I write for — Grok at $15/month during the promo is honestly a good deal. It’s not the best model. But it might be the best value right now.
The Bottom Line
The Musk-Altman feud is entertainment. The tools are what matter. And right now, Grok 4.5 is a legitimate option that costs half what its competitors charge. It’s fast, it’s forgiving, and it’s good enough for most everyday tasks. The hallucination issue is real — verify anything factual. But at $15/month through July 15, it’s worth testing for yourself.
If you’re new to AI tools entirely, start with my guide on the AI tools I actually use every day — it’ll help you figure out what you need before you commit to any subscription. And if you want help picking the right tool for your specific situation, check the AI Tool Advisor.