xAI made Grok 4.6 available in Amazon Bedrock
Grok 4.6 from xAI has been available in Amazon Bedrock since 18 August 2026, with a context window of 500 thousand tokens and four reasoning effort levels, targeting long-running agents and work with code.
xAI made Grok 4.6 available in Amazon Bedrock on 18 August 2026 — it is the second model from xAI on this platform after Grok 4.3. The model offers a context window of 500 thousand tokens and four adjustable reasoning effort levels: low, medium, high and xhigh. It is available on both bedrock-mantle and bedrock-runtime endpoints and supports Converse API (including streaming), Chat Completions and Responses. Two cross-region inference profiles are available: us.xai.grok-4.6 for operation within the USA (for data residency requirements) and global.xai.grok-4.6 for worldwide routing with greater capacity.
According to xAI, Grok 4.6 is designed for long-running agents, work with codebases and knowledge work, and builds on Grok 4.5. The company states that the model underwent longer additional training with curated data covering advanced technical concepts and engineering data, and was trained on a broad range of agentic reinforcement learning tasks, including kernel optimization, web development and CAD. The company further claims that the model more often tests and verifies its own work on longer tasks and delivers stronger results on the first attempt for visual and interactive projects. Regarding safety, the company states that the mechanisms were calibrated to the capabilities of the model and underwent what it describes as the most extensive suite of tests before and after deployment, including third-party testing.
When releasing the model on 12 August 2026, xAI published results for the Grok 4.6 High variant in several benchmarks: AA Intelligence Index 61, GDPVal-AA v2 1753, CursorBench v3.2 69.9 %, DeepSWE v1.1 65.9 %, FrontierCode v1.1 61.3 %, APEX-Agents 57.5 %, Terminal-Bench v3.0 26 %, APEX-SWE 56.4 %, AA-Briefcase 1577 and Harvey LAB (Vals) 15.8 %. According to the source, some of these benchmarks (including AA Intelligence Index and AA-Briefcase) come from Artificial Analysis and measure agentic tool use, knowledge, reliability and analytical work with documents.
You can find details in the source article.
Why it matters
For developers working with AWS, this expands the range of models in Bedrock with an alternative to models from OpenAI and Anthropic focused on long agentic tasks and work with code, accessible through the standard Bedrock interface (Converse API) rather than only through an OpenAI-compatible client. The choice between regional (USA) and global routing profiles allows companies to account for data residency requirements when deploying agents.
Release card
Grok 4.6
xAI
- Context
- 500K token context window (500 thousand tokens)
- AA Intelligence Index 61 An aggregate intelligence index from Artificial Analysis combining nine evaluations of agentic tool use, knowledge and analytical work.
- GDPVal-AA v2 1753 One of the nine individual evaluations included in AA Intelligence Index.
- CursorBench v3.2 69,9 % The result for the Grok 4.6 High variant in this benchmark, according to xAI.
- DeepSWE v1.1 65,9 % The result for the Grok 4.6 High variant in this benchmark, according to xAI.
- FrontierCode v1.1 (Extended) 61,3 % The result for the Grok 4.6 High variant in this benchmark, according to xAI.
- Long-running agentic tasks spanning many steps, such as research, analysis or work with a codebase
- Visual and interactive projects where, according to xAI, the model delivers stronger results on the first attempt
- Agentic coding and engineering tasks, such as kernel optimization, web development or CAD
Builds on Grok 4.5 from xAI, with longer additional training and a broader range of agentic reinforcement learning tasks focused on long-running agents and more demanding interactive and visual tasks.
The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →
Two audiences, two different impacts
What this means
For individuals
Developers building agents or tools for working with code on AWS can now try Grok 4.6 in Bedrock through Converse API with streaming and four reasoning effort levels, without needing an OpenAI-compatible client.
For a business
Companies running agentic or coding tools on AWS Bedrock gain another model option, with a choice of regional (USA) or global routing depending on data residency requirements and available capacity.
DevelopmentCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.