Grok 4 scored 57% in “The Last Exam” versus 22% for Gemini 2.5 Pro

Elon Musk presented a new version of his neural network – Grok 4. The maximum version – Grok 4 Heavy – can run multiple computations simultaneously and scores 57% in the most difficult test “The Last Exam of Humanity”. For comparison, the previous leader Gemini 2.5 Pro showed only 22%.

The model has become better at understanding videos and images. Voice mode has improved. Now Grok can not only speak, but also whisper, sing, change intonations and emotions. In the medical benchmark X-ray, Grok 4 shows 72.8%. This is higher than all other models.

According to independent assessment by Artificial Analysis experts, Grok 4 is now truly the smartest model, surpassing even o3-pro from OpenAI.

The model can already assemble 3D games: finds assets itself, selects textures and adds 3D models to the engine. The context window in the API is 256,000 tokens.

In the vending benchmark, Grok 4 earned $4694 and sold more than 4,000 items. For comparison, the new Claude Opus could only earn $2077.

The maximum version Grok 4 Heavy will be available only by subscription for $300 per month. The basic plan for $30 includes regular Grok 4, and free users work with Grok 3 for now.

In August, the release of a separate coding model is expected, in September a multimodal agent will appear. And in October a version for video generation will be released. All these components will become part of the Grok 4 ecosystem.

AIvengo > Reviews > Grok 4 scored 57% in “The Last Exam” versus 22% for Gemini 2.5 Pro

UBTech will send Walker S2 robots to serve on China's border for $37 million

Chinese company UBTech won a contract for $37 million. And will send humanoid robots Walker S2 to serve on China's border with Vietnam. South China Morning Post reports that the robots will interact with tourists and staff, perform logistics operations, inspect cargo and patrol the area. And characteristically — they can independently change their battery.

Anthropic accidentally revealed an internal document about Claude's "soul"

Anthropic accidentally revealed the "soul" of artificial intelligence to a user. And this is not a metaphor. This is a quite specific internal document.

Jensen Huang ordered Nvidia employees to use AI everywhere

Jensen Huang announced total mobilization under the banner of artificial intelligence inside Nvidia. And this is no longer a recommendation. This is a requirement.

AI chatbots generate content that exacerbates eating disorders

A joint study by Stanford University and the Center for Democracy and Technology showed a disturbing picture. Chatbots with artificial intelligence pose a serious risk to people with eating disorders. Scientists warn that neural networks hand out harmful advice about diets. They suggest ways to hide the disorder and generate "inspiring weight loss content" that worsens the problem.

OpenAGI released the Lux model that overtakes Google and OpenAI

Startup OpenAGI released the Lux model for computer control and claims this is a breakthrough. According to benchmarks, the model overtakes analogues from Google, OpenAI and Anthropic by a whole generation. Moreover, it works faster. About 1 second per step instead of 3 seconds for competitors. And 10 times cheaper in cost per processing 1 token.