The new AI model improves performance in programming and long-running tasks, with a price starting at USD 2 per million input tokens and USD 6 per million output tokens
The private equity firm SpaceXAI (under the stock ticker SPCX) shook the foundations of the global industry by officially launching its cutting-edge model: Grok 4.6 in the Artificial Intelligence market.
This demonstration of strength and private innovation immediately propelled the company's stock by a spectacular +11.17%, trading at a solid value of 148.184 dollars per share, confirming that investors reward the company's boldness against the bureaucratic stagnation of its competitors.
Elon Musk
The model positions itself from the outset as the gold standard in productive efficiency, specifically designed to optimize long-execution agents, complex programming, and the most demanding scenarios in knowledge work.
The new system completely redefines technical execution by focusing on high-duration multiphase tasks that traditional models, burdened by inefficiencies, cannot sustain.
The evolution from Grok 4.5 to Grok 4.6 translates into an unprecedented capacity for thematic research, automation of information analysis, collaboration across multiple code repositories, and the immediate conversion of business ideas into functional and monetizable applications.
Official spokespersons for the company proudly highlighted that "the new model demonstrates superior self-assessment and verification capabilities in long-execution tasks".
The scientific superiority of private management is reinforced by empirical performance data. In the demanding AI Intelligence Analysis Index (an indicator that thoroughly evaluates overall capability through a battery of 9 cutting-edge benchmarks), the elite version Grok 4.6 High achieved the extraordinary score of 61 points.
SpaceXAI
With this record figure, the creation of the private equity firm ties for the absolute technological throne with the corporate giant GPT-5.6 Sol Max (also referred to as GPT-5.6 Sol), far surpassing the 56 points achieved by the previous model Grok 4.5 High.
A detailed and transparent breakdown of the technical metrics shows that the xAI / SpaceXAI model humiliates its slower design competitors in the vast majority of critical industry tests:
Harvey LAB (The demanding benchmark focused on legal agents): In one of the most abysmal differences ever recorded, Grok 4.6 achieved a crushing 15.8% accuracy, pulverizing the ridiculous 2.5% obtained by GPT-5.6 Sol and far exceeding the 11.3% of Fable 5.
GDPVal-AA v2 (Evaluation of complex reasoning and data validation): The free enterprise model crowned itself absolute leader with 1,753 points, surpassing GPT-5.6 Sol (1,728 points) and Fable 5 (1,741 points).
AA-Briefcase (Productivity and information management metric for business agents): Grok 4.6 led the table with 1,577 points, exceeding the 1,502 of GPT-5.6 Sol and the 1,574 of Fable 5.
CursorBench v3.2 (Efficiency in assisted code development environments): Achieved an outstanding 69.9%, placing above the 67.2% of GPT-5.6 Sol, and just below the 70.5% recorded by Fable 5.
APEX-Agents (Operational effectiveness of autonomous agents): Reached a 57.5%, beating the 56.7% of GPT-5.6 Sol (though below the 59.2% of Fable 5).
APEX-SWE (Agent-based software engineering development): Recorded a 56.4%, surpassing the 53.6% of its predecessor, in a table closely led by Fable 5 with 58.8% (with no records available for GPT-5.6 Sol).
FrontierCode v1.1 Extended (Advanced programming metric in frontier environments): Achieved a 61.3%, beating the 60.6% of GPT-5.6 Sol, slightly below Fable 5 with 63.6%.
DeepSWE v1.1 (Deep software engineering environments): Achieved a solid 65.9% performance, compared to the 73% of GPT-5.6 Sol and the 70% of Fable 5.
Terminal-Bench v3.0 (Direct interaction with command terminals): Obtained a 26%, falling behind the 34.6% of GPT-5.6 Sol and the 34.1% of Fable 5.
Grok 4.6
This miracle of modern engineering is the direct result of an optimization process guided by economic rationality and high-efficiency capital investment.
The model underwent an extended phase of supplemental training, using high-quality data algorithmically curated for advanced technical reasoning. xAI brilliantly utilized Grok 4.5 to regenerate supervised fine-tuning trajectories (SFT) in key productivity areas such as STEM disciplines (science, technology, engineering, and mathematics), software development, and knowledge work, rigorously filtering any problematic trajectory through automatic validations. Subsequently, the model completed its maturation with aggressive agent-based reinforcement learning (RL) training focused on web development, computer-aided design, kernel optimization, and general programming.
As a result of this exquisite architecture, the model is capable of structuring the code and visual language of complete applications "in a single pass".
True to the principles of deregulation and commercial ease that characterize companies in the SpaceXAI ecosystem, the model's availability is immediate and globally accessible.
It is integrated from today into competitive development platforms such as Cursor, Grok Build, the company's official API, as well as in market aggregators OpenRouter, Vercel, and Cloudflare.
Grok 4.6
The pricing structure is a monument to cost efficiency and accessibility for entrepreneurs: the standard version costs just 2 dollars per million input tokens and 6 dollars per million output tokens, while the high-speed version is offered at exactly double that commercial price. Additionally, to celebrate this triumph of technological capitalism, the company will grant double usage credits for free on Grok Build and Cursor throughout its first week on the market.
Finally, in the vital area of security, the company has avoided state regulatory paralysis and implemented safeguards calibrated with mathematical precision to maximize productive utility and the sovereignty of the legitimate user.
The robustness of the security stack was confirmed after the most rigorous pre- and post-deployment testing battery in history, evaluated both internally and by independent external auditors, demonstrating superlative capabilities in key areas for business protection such as self-correction of code vulnerabilities
