SpaceXAI releases Grok 4.6 to take on GPT-5.6 Sol in agentic AI

SpaceXAI releases Grok 4.6 to take on GPT-5.6 Sol in agentic AI

Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index while improving performance across coding, knowledge work, and long running agent tasks.

Share

Add us on Google
By Editorial Team
Edited by Estefano Gomez

SpaceXAI has released Grok 4.6, its latest artificial intelligence model, with improvements focused on long running agents, coding, knowledge work, and interactive visual projects.

The company said Grok 4.6 builds on Grok 4.5 and is designed to remain effective across complex tasks involving many steps, including research, information analysis, software development, and building applications.

Grok 4.6 scored 61 on the Artificial Analysis Intelligence Index, matching OpenAI’s GPT-5.6 Sol Max and trailing Fable 5 Max at 62. Grok 4.5 High scored 56 on the same benchmark.

The model also outperformed GPT-5.6 Sol Max on several evaluations published by SpaceXAI. Grok 4.6 scored 1,753 on GDPVal-AA v2 compared with 1,728 for GPT-5.6 Sol Max, while CursorBench v3.2 results came in at 69.9% versus 67.2%.

Advertisement

On FrontierCode v1.1 Extended, Grok 4.6 scored 61.3% compared with 60.6% for GPT-5.6 Sol Max. It also scored 57.5% on APEX Agents versus 56.7% for GPT-5.6 Sol Max.

GPT-5.6 Sol Max remained ahead on DeepSWE v1.1, scoring 73% compared with Grok 4.6 at 65.9%, and on Terminal Bench v3.0, where it scored 34.6% versus 26%.

SpaceXAI said Grok 4.6 underwent a longer supplemental training run than Grok 4.5 using curated model generated data, engineering data, and changes to its optimizer and training process. 

Grok 4.5 was also used to regenerate supervised fine tuning trajectories across reasoning, software engineering, STEM, and knowledge work.

The company subsequently trained the model through reinforcement learning across agentic tasks covering general coding, kernel optimization, web development, computer aided design, and other environments.

SpaceXAI said testing showed Grok 4.6 was particularly capable of turning broad product ideas into working initial versions. The model can research unfamiliar areas, structure applications, implement interactions, and refine its output across multiple rounds of feedback.

The company also reported increased self testing and verification during longer tasks, along with stronger initial results for visual and interactive projects compared with Grok 4.5.

Grok 4.6 is available through Cursor, Grok Build, the SpaceXAI API, OpenRouter, Vercel, and Cloudflare.

API pricing starts at $2 per million input tokens and $6 per million output tokens. A faster version is available at twice the price.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
AI
SpaceXAI releases Grok 4.6 to take on GPT-5.6 Sol in agentic AI
SpaceXAI releases Grok 4.6 to take on GPT-5.6 Sol in agentic AI

Grok 4.6 matches GPT-5.6 Sol on the Artificial Analysis Intelligence Index while improving performance across coding, knowledge work, and long running agent tasks.

By Editorial Team
Edited by Estefano Gomez

Share

Add us on Google

SpaceXAI has released Grok 4.6, its latest artificial intelligence model, with improvements focused on long running agents, coding, knowledge work, and interactive visual projects.

The company said Grok 4.6 builds on Grok 4.5 and is designed to remain effective across complex tasks involving many steps, including research, information analysis, software development, and building applications.

Grok 4.6 scored 61 on the Artificial Analysis Intelligence Index, matching OpenAI’s GPT-5.6 Sol Max and trailing Fable 5 Max at 62. Grok 4.5 High scored 56 on the same benchmark.

The model also outperformed GPT-5.6 Sol Max on several evaluations published by SpaceXAI. Grok 4.6 scored 1,753 on GDPVal-AA v2 compared with 1,728 for GPT-5.6 Sol Max, while CursorBench v3.2 results came in at 69.9% versus 67.2%.

Advertisement

On FrontierCode v1.1 Extended, Grok 4.6 scored 61.3% compared with 60.6% for GPT-5.6 Sol Max. It also scored 57.5% on APEX Agents versus 56.7% for GPT-5.6 Sol Max.

GPT-5.6 Sol Max remained ahead on DeepSWE v1.1, scoring 73% compared with Grok 4.6 at 65.9%, and on Terminal Bench v3.0, where it scored 34.6% versus 26%.

SpaceXAI said Grok 4.6 underwent a longer supplemental training run than Grok 4.5 using curated model generated data, engineering data, and changes to its optimizer and training process. 

Grok 4.5 was also used to regenerate supervised fine tuning trajectories across reasoning, software engineering, STEM, and knowledge work.

The company subsequently trained the model through reinforcement learning across agentic tasks covering general coding, kernel optimization, web development, computer aided design, and other environments.

SpaceXAI said testing showed Grok 4.6 was particularly capable of turning broad product ideas into working initial versions. The model can research unfamiliar areas, structure applications, implement interactions, and refine its output across multiple rounds of feedback.

The company also reported increased self testing and verification during longer tasks, along with stronger initial results for visual and interactive projects compared with Grok 4.5.

Grok 4.6 is available through Cursor, Grok Build, the SpaceXAI API, OpenRouter, Vercel, and Cloudflare.

API pricing starts at $2 per million input tokens and $6 per million output tokens. A faster version is available at twice the price.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.