-
2026-07-16
Grok 4.5 released
xAI released Grok 4.5, trained using hundreds of thousands of NVIDIA GB300 GPUs, supporting asynchronous execution and multi-step software engineering tasks.
-
2026-07-22
Grok 4.5 launched across multiple platforms
Grok 4.5 officially released on iOS, Android, Web, and X platforms, supporting long-duration dialogue tracking and independent completion of complex tasks.
-
2026-08-12
Grok 4.6 released
xAI announced the launch of Grok 4.6, focusing on long-running agents, interactive, and visual work based on Grok 4.5.
-
2026-08-19
Grok 4.6 launched on Amazon Cloud
Grok 4.6 officially launched through Amazon Bedrock, offering configurable inference intensity options.
-
2026-08-26
Grok 4.6 launched on Microsoft Cloud
Grok 4.6 was launched through Microsoft Foundry, designed for long-running agents and interactive visual work.
↗
On July 16, 2026, SpaceXAI released its most powerful model, Grok 4.5, designed specifically for programming, agent-based tasks, and knowledge-based work. Grok 4.5 was trained on datasets in the fields of coding, science, engineering, and mathematics, and performed better than comparable leading models in real-world engineering tasks. The model was trained using hundreds of thousands of NVIDIA GB300 GPUs and employs reinforcement learning techniques. It supports asynchronous execution and can handle multi-step software engineering tasks. Grok 4.5 provides services at a rate of 80 requests per second, with token efficiency twice that of the latest models in its category. The cost is $2/$6 per million input/output tokens. Currently, Grok 4.5 has become the default model for Grok Build, enabling efficient construction of complex Excel models, PowerPoint charts, and Word document content.
↗
xAI announced that its latest model, Grok 4.5, is now available on iOS, Android, Web, and X platforms. As the most intelligent version to date, Grok 4.5 has been officially launched on grok.com, X, and corresponding mobile apps. In daily use, Grok 4.5 can more accurately follow user instructions, track long conversations, and provide clear and reliable responses; when handling complex tasks, it can independently complete tasks from start to finish, such as creating spreadsheets, generating slides, or extracting key points from PDFs. Additionally, the model runs directly in Word, Excel, PowerPoint, and Outlook through Microsoft 365 plugins.
↗
# Introducing Grok 4.6
Grok 4.6 builds on Grok 4.5 with a focus on long-running agents, more ambitious interactive and visual work, and frontier intelligence across coding and knowledge work benchmarks. It achieves frontier-level performance on the Artificial Analysis Intelligence Index, matching GPT-5.6 Sol on the composite score of nine benchmarks.
Grok…
↗
# Grok 4.6 on Amazon Bedrock
Grok 4.6 is now available via Amazon Bedrock. Today, we're excited to announce that Grok 4.6 is now generally available on Amazon Bedrock. Grok 4.6 is our latest flagship model, built for long-running agents and ambitious interactive and visual work. It offers 500k context window and configurable reasoning efforts (low, medium,…
↗
On August 26, 2026, XAI announced that its flagship model Grok 4.6 has been launched through Microsoft Foundry. This model is designed for long-term running agents and interactive visual work, offering a context window with 500,000 words and configurable reasoning intensity levels (low, medium, high, xhigh). Microsoft Foundry allows organizations to evaluate the performance of Grok 4.6 against other advanced models on a single platform, conduct specific workload tests, deploy managed endpoints, and implement enterprise-level security and governance controls. Developers can obtain deployment details for building coding agents, engineering assistants, or enterprise automation from the Foundry model catalog.