Grok vs Claude: Claude leads by 10 points at a similar price

Which is better, Grok or Claude?

Claude is better on measured quality, and the gap isn't small. On the Artificial Analysis Intelligence Index (version 4.3.2, read on 11 October 2026), Claude Sonnet 5.5 scores 56 and Grok 4.7 scores 46. Both models came out in September 2026.

Price doesn't explain the gap. Artificial Analysis blends each price at a 7:2:1 mix of cached input, fresh input and output, which gives $1.47 per million tokens for Claude Sonnet and $1.35 for Grok. So this is a fair fight. xAI calls Grok 4.7 "our most capable model for coding and knowledge work" in its launch post, and it still trails Anthropic's middle tier by 10 points. Against Claude Opus 5.5, the model Anthropic tells developers to start with, the gap is 12. Grok has no bigger model in this data: xAI's model docs list nothing above Grok 4.7, and Artificial Analysis has no entry for Grok's Heavy mode, which grok.com's subscription page describes as a "Team of Experts" running Grok 4.7, so that mode isn't measured here.

Claude leads on 9 of the 10 tests inside the index. The tenth is a tie. On AA-Omniscience, which scores what a model knows against how often it makes things up, both land on 32. And Grok has a feature no Claude plan lists: xAI's Grok page says it "Searches the web and 𝕏 live", so it can tell you what people posted on X an hour ago.

How do the current Grok and Claude models compare?

Anthropic sells four current models. xAI's docs recommend one, Grok 4.7, and grok.com's subscription page names it for both its Expert and Heavy modes. Grok 4.6, the model before it, is listed for reference. Here they are side by side, every figure read from Artificial Analysis on 11 October 2026 at the highest effort it tested, with context windows from each vendor's model docs:

Anthropic's own advice is short. Its models overview says: "If you're unsure which model to use, start with Claude Opus 5.5 for most workloads." xAI's model docs point to Grok 4.7 for "everything else, including code".

Claude is also faster in this data. Every Claude model streams output quicker than Grok 4.7 except Fable, the slowest at 65.3 tokens a second. Claude Haiku runs at 235.0.

Effort settings change the picture a little. At high effort, Grok 4.7 scores 46, and it costs $2.73 for each task in the index. Claude Sonnet 5.5 at the same high setting scores 47, and it does that for $0.88. That's the same score for about a third of the money. Grok does win on the raw output price, which is $6 per million output tokens to Sonnet's $10. But its cost per task still came out higher in these runs, so the cheap tokens didn't make it the cheap model.

Context is the other gap. Every current Claude model can take in 1M tokens, and it can write up to 128K of them in one reply. Grok 4.7 takes 500K. On top of that, xAI bills a prompt of 200K tokens or more "at the higher rate for all tokens in the request", which is $4 in and $12 out per million, or double the base rate.

How much do Grok and Claude Pro cost?

Claude Pro costs $20 a month and Grok Premium costs $30, so Claude is the cheaper of the two main paid plans by $10. On 12 October 2026 xAI moved new customers from SuperGrok to The X Subscription, and Premium is its replacement for SuperGrok. Both labs sell a $100 tier and a $200 tier, and only xAI sells a monthly plan below $20. US prices, checked on 26 October 2026 on claude.com/pricing, x.ai/pricing and xAI's launch post:

Neither lab prints a message count. Anthropic's Max help page defines its tiers as multiples of Pro, and xAI sells Premium, Super and Ultra as 4x, 15x and 50x the usage of Plus, in one weekly pool shared across Grok and Cursor. X's help centre still lists Premium+ at $40 a month in the US, with "SuperGrok access", as of October 2026, while xAI's docs say Premium and the tiers above it carry "full X Premium+ benefits". Is SuperGrok worth it works through the X bundle, and Claude Pro vs Claude Max does the same for Anthropic's upper tiers.

If you already pay for SuperGrok, you don't have to move. xAI's launch post says "Nothing changes for existing SuperGrok, Cursor, or X Premium subscribers", and if you do switch you'll "get a prorated refund for your old plan". One catch on the new plans: xAI says Claude and other third party models in Cursor are "billed separately" at API list prices.

What does Grok do that Claude does not?

Grok searches X and makes pictures and video. No Claude plan lists either one. Anthropic's docs say all current Claude models "support text and image input, text output", so Claude can look at an image but can't make one.

Live X search. xAI's Grok page says it "Searches the web and 𝕏 live", and the Free tier gets "Daily access to Grok". Claude's plans list web search, but they don't list X search. When a launch, a crisis or a market move is only an hour old, Grok can see the posts before most of the articles about it exist.

Grok Imagine. xAI's Grok page lists images "Up to 2K resolution" and "Text-to-video up to 15 seconds at 720p", and its pricing page lists "Make stunning AI images & videos" on Premium at $30. There is no Claude plan that makes images or video at all.

Grok Bot. xAI's pricing page lists Grok Bot from Premium up. The closest thing on Claude's side is the "Hand off and schedule tasks" line on Pro.

The X bundle. Premium, Super and Ultra carry "full X Premium+ benefits", per xAI's docs. And if you already pay for X Premium+ on X, its help centre says that includes SuperGrok access, so you don't need to buy Grok again. Nobody bundles Claude with anything like that.

Is Claude or Grok better for coding and long documents?

Claude is better for both. The coding gap is the biggest one in the data: Claude Sonnet 5.5 scores 64% on Terminal-Bench 4.0, Artificial Analysis's test of agentic coding in a terminal, and Grok 4.7 scores 26%. On SciCode, a scientific coding test, it's closer, at 61% to 57%.

The tools follow the scores. Claude Code comes with Pro at $20 and with both Max tiers, but Anthropic doesn't list it on Claude Free. xAI's coding agent is Grok Build, and Grok 4.7 launched in Grok Build, Cursor and the API. If you code with an agent every day, Claude is where the measured results are.

Long documents go the same way. Claude takes 1M tokens to Grok 4.7's 500K, and on AA-LCR, the index's long context reasoning test, Claude Sonnet scores 83% to Grok's 77%. Projects, which keep files and instructions together across chats, are on every Claude plan, five of them on Free.

For writing there's no clean benchmark, so we won't invent one. What the data shows is that Claude leads on the reasoning tests and ties on knowledge. Tone is a taste call, and you should run your own draft through both before you pay. Grok vs ChatGPT covers Grok's looser guardrails in more depth.

Do Grok and Claude train on your chats?

Both can, and both let you say no. Neither vendor's page we read states which way the switch starts, so check yours. The settings and retention rules, from each company's own pages as of October 2026:

The Claude rules come from Anthropic's consumer terms update and its privacy centre. The Grok rules come from xAI's consumer FAQ and privacy policy. Anthropic's privacy centre says it keeps the feedback you send with the thumbs up or down buttons for up to five years, so skip the rating on anything you want gone.

Should you pick Grok or Claude?

Pick Claude unless your work depends on X or on video. Here's the split by situation:

If you want both models without two bills, Whizi Pro at $29.99 a month puts Claude Opus, Claude Sonnet and Grok in one chat, with a side by side compare for the same prompt. A Grok message costs 6 credits, Claude Sonnet 10 and Claude Opus 20, out of 2,500 a month. It doesn't carry live X search, Grok Imagine video or Claude Code, so if one of those is why you're buying, the vendor plan is the right one.