@FinanceYF5: 1/ Grok 4.6 just released. Same price as 4.5. Stronger agents. Better coding. Runs longer on complex tasks. AA Intelligence Index ties GPT-5.6 Sol. $2 per million input tokens, $6 per million output tokens.
Summary
Grok 4.6 is officially released, with the same pricing as 4.5, enhanced agent and programming capabilities, and it runs longer on complex tasks. The AA Intelligence Index ties GPT-5.6 Sol, with input at $2 per million tokens and output at $6 per million tokens.
View Cached Full Text
Cached at: 08/13/26, 09:24 PM
1/ Grok 4.6 just released.
Price same as 4.5.
Agent stronger. Programming better. Complex tasks run longer.
AA Intelligence Index catches up to GPT-5.6 Sol.
$2 per million input Tokens, $6 per million output Tokens. 👇 https://t.co/jmdP1F8DUH
2/ The focus of this upgrade is not “answering more intelligently.”
It’s about letting Agents work continuously for longer.
Research unfamiliar domains, analyze materials, modify entire codebases, and turn an idea into a usable app or work product.
3/ Give Grok 4.6 a product idea, and it can research on its own, build the structure, complete core interactions, and then continue modifying based on feedback.
When a task runs long enough, it will also proactively test and check its own work.
Agents are starting to resemble real project executors.
4/ Overall capability:
Fable 5: 62 Grok 4.6: 61 GPT-5.6 Sol: 61 Grok 4.5: 56
But in the GDPVal-AA knowledge-work benchmark, Grok 4.6 scored 1753, surpassing Fable 5 and GPT-5.6 Sol.
5/ Coding ability also improved significantly compared to 4.5:
DeepSWE: 54% → 65.9% CursorBench: 66.7% → 69.9% FrontierCode: 56.6% → 61.3%
But it’s not #1 across the board; GPT-5.6 Sol and Fable 5 still lead in some coding benchmarks.
6/ Behind this is a longer supplementary training run than Grok 4.5.
xAI added reasoning, engineering, and advanced technical data, and had Grok 4.5 regenerate SFT trajectories.
Reinforcement learning covered tasks like knowledge work, general programming, kernel optimization, web development, and CAD.
7/ Grok 4.6 is now available on Cursor, Grok Build, API, OpenRouter, Vercel, and Cloudflare.
The standard version’s price stays the same, while the Fast version is priced at double.
Cursor and Grok Build offer 2x included usage for the first week; you can try it right now.
Official release:
Grok Bot has taken the entire internet by storm, and for good reason.
It’s a very excellent AI Agent.
But only if you configure it the right way.
These 5 steps can make your Grok Bot Agent team extremely powerful.
Similar Articles
Grok 4.6
xAI releases Grok 4.6, a frontier model focused on long-running agents and ambitious interactive/visual work, matching GPT-5.6 Sol on the Artificial Analysis Intelligence Index and available in Cursor and Grok Build.
SpaceXAI releases Grok 4.5, which Elon describes as an ‘Opus-class model’
SpaceXAI has released Grok 4.5, an Opus-class model that offers faster speed, higher token efficiency, and lower cost compared to competitors like Anthropic's Opus 4.7 and OpenAI's models.
Grok-4.5 on par with gpt-5.5-xhigh in coding at half the cost
Grok-4.5 achieves coding performance comparable to GPT-5.5-xhigh while costing half as much.
@interjc: 我冲 Premium+ 送的 SuperGrok 价值还在上升?
用户感叹开通 Premium+ 赠送的 SuperGrok 价值上升,引用 Elon Musk 宣布 Grok 4.6 发布,称其智能、快速且性价比高。
@mckaywrigley: grok 4.6 is great. its intelligence per dollar is crazy good. fable 5 is still the best model (and i don’t think it’s c…
The author praises Grok 4.6 for its strong intelligence per dollar, declares Fable 5 still the best model, and says he is intrigued by Grok 4.7, claiming SpaceX/Cursor is now firmly a frontier AI player alongside Anthropic and OpenAI.