Grok 4.6 Launches With Benchmarks to Back It
SpaceXAI's Grok 4.6 is drawing heavy chatter backed by real throughput and benchmark numbers, not just launch hype
What's happening
SpaceXAI shipped Grok 4.6, pitched for long-running agent work with pricing held steady, and independent testers cited concrete numbers: 85.8 tokens/second versus a 71.2 baseline, a 61 score on the Artificial Analysis Intelligence Index, and reasoning benchmarks matching GPT-5.6 Sol at roughly one-fifth the output cost. Grok is also being positioned as an AI "teammate" that needs account access to do work.
The numbers
Recent mentions hit 65 — the highest raw volume in this batch — with velocity 7.0 across Hacker News, Mastodon and The Verge. No market price applies; SpaceXAI/xAI isn't a traded security.
Our read
Verdict: watch. The chatter is backed by specific, repeatable benchmark numbers from more than one source, a stronger signal than pure launch-day hype. Worth tracking adoption of the "AI teammate" account-access model given the privacy pushback visible in the same chatter.
Sources we checked
- Elon Musk seems to think that asshole behaviour is all about freedom of speech. I saw this today about grok:My immediate thoughts were1. Wow Grok is an SpaceX p — @dermdaly@mastodon.social
- Grok 4.6's throughput clears median for its tier—85.8 tokens per second versus 71.2 baseline. But pricing doubles past 200K prompt tokens. Watch whether agents — @schuler@mastodon.social
- Grok 4.6 matches GPT-5.6 Sol's reasoning benchmark at one-fifth the output cost, per independent testing. For agents running many prompts before completion, the — @schuler@mastodon.social
- 「Grok 4.6」が発表 ~長時間稼働するエージェント作業に特化、価格は据え置き/「GPT-5.6 Sol」に並ぶ知能スコア。「Cursor」「Grok Build」で即日利用可能https://forest.watch.impress.co.jp/docs/news/2132441.html#forest_watc — @forest_watch_impress@rss-mstdn.studiofreesia.com
- Grok 4.6 — Hacker News
- SpaceXAI Wants Grok Bot to Do Your Job—But It Needs Access to Your Accounts — decrypt
- SpaceXAI's Grok 4.6 Scores 61 on the Artificial Analysis Intelligence Index — Hacker News
- i burned 4 openai resets, burned all my kimi k3 credits, put some extra credits into deepseek, burned all grok credits in one day.. 😂 needed a 2nd openai sub, h — @michabbb@vivaldi.net
See today's live board
This story is frozen as published on August 13, 2026. The live board shows what is moving right now, scored against real money.
Open this story live → See today's live boardMore AI stories
- Rogue Agent Hack Meets OpenAI Leadership Exodus August 15, 2026
- Anthropic's IPO buildup collides with a bug-bounty and legal win August 15, 2026
- OpenAI's rogue agent hacked five platforms, not one August 14, 2026
- DeepSeek Hikes API Prices Up to 1000% August 13, 2026
- DeepSeek Ships V4 Pro, Undercuts On Price August 13, 2026
- OpenAI Slows Astra Over Cyber Risk, Not Hype August 8, 2026
Research, not financial advice. Sort Brick reports what its data showed at the time of publication. It is not a recommendation to buy or sell anything — you decide.