August 12, 2026
Benchmarks, beef, and bot chaos
Grok 4.6
Grok 4.6 drops and the comments instantly turn into a benchmark brag war
TLDR: xAI launched Grok 4.6, saying it’s better at long tasks, building projects, and visual work while matching top rivals on major scoreboards. Commenters instantly turned it into drama, debating whether the timing was a strategic shot at competitors and cheering the speed-and-price flex.
xAI says Grok 4.6 is here, promising a smarter chatbot that can stick with long projects, build apps from rough ideas, handle more visual work, and even check its own work as it goes. It’s launching across Cursor, Grok Build, the API, and partners like OpenRouter, Vercel, and Cloudflare, with a one-week 2x usage promo clearly designed to get people clicking fast. The company is also flexing hard on pricing, saying it starts cheap enough to tempt developers who are comparison-shopping every new model release.
But the real action is in the comments, where the mood is basically: “Okay, but is this a power move?” One of the first reactions immediately clocked the timing, asking if the launch was suspiciously close to DeepSeek’s latest release. Translation for normal humans: the AI companies are dropping products like pop stars releasing diss tracks. Others skipped the conspiracy and went straight into full hype mode, calling it “Fable level performance” and shouting about it being faster and significantly cheaper. Another commenter piled on with the benchmark chest-thumping, saying it beats GPT-5.6 Sol on most tests and undercuts rivals on price.
The funniest part? Nobody in this tiny thread is acting calm. It’s all vibes: benchmark bragging, price-war glee, and launch-timing side-eye. Even the plain link-drop to Cursor reads like someone entering the group chat with receipts. In other words, Grok 4.6 didn’t just launch a model — it launched a fresh round of AI leaderboard drama.
Key Points
- •Grok 4.6 was released as an update to Grok 4.5 with a focus on long-running agents and interactive and visual project work.
- •The article says Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index, which combines nine benchmarks.
- •Training for Grok 4.6 included a longer supplemental run, curated model-generated reasoning data, engineering data, improved optimization, and further SFT and RL stages.
- •The article reports stronger performance on multi-step product-building tasks, including self-testing, verification, and better first-pass visual and interactive outputs than Grok 4.5.
- •Grok 4.6 is available in Cursor, Grok Build, the API, and through OpenRouter, Vercel, and Cloudflare, with pricing starting at $2 per million input tokens and $6 per million output tokens.