![\"tech\"](\"https:\/\/images.firstpost.com\/dlxczavtqcctuei\/news18\/static\/images\/fp\/revamp_v1\/tech.svg\")
More from Tech<\/p>\nSpaceX AI has unveiled Grok 4.6, its latest model aimed at advanced coding, agentic tasks and knowledge work, with the company positioning it as a rival to leading models from OpenAI and Anthropic. <\/p>\n
SpaceX AI seems to be on a roll, with the company adding new AI tools and models to its portfolio at a rapid pace. Following the recent acquisition of Cursor and the release of the Grok Bot yesterday, the Elon Musk-led company is stepping up its efforts to compete with other leading AI models. Building on that push, SpaceX AI has unveiled Grok 4.6, claiming that its intelligence is comparable to OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable 5.<\/p>\n
Grok 4.6 has been designed for long-running agents, coding, knowledge work and more ambitious visual projects. The company says the model can stay on task across multiple steps, work through unfamiliar codebases and increasingly test its own output before moving forward.<\/p>\n
How Grok 4.6 has been trained<\/p>\n
According to SpaceX AI, Grok 4.6 underwent a longer training run than Grok 4.5, using curated model-generated data focused on reasoning and advanced technical concepts, high-quality engineering data, an improved optimiser and a refined training process.<\/p>\n
Throughout training, Grok 4.5 was used to regenerate supervised fine-tuning (SFT) trajectories across reasoning tasks, agent harnesses and domains such as STEM, software engineering and knowledge work. Problematic traces were filtered out using model-based checks. The resulting SFT checkpoint demonstrated stronger performance and improved behaviour.<\/p>\n
More from Tech<\/p>\n
The model was also trained on a wide range of agentic reinforcement-learning tasks, including knowledge work, general coding and domain-specific environments for kernel optimisation, web development, computer-aided design and more.<\/p>\n
Safety guardrails<\/p>\n
SpaceX AI says Grok 4.6’s safeguards have been improved and calibrated to match the model’s capabilities. The safety stack is designed to maximise both utility and security across legitimate use cases, allowing Grok 4.6 to assist with tasks such as vulnerability patching, accelerating engineering design and supporting AI research.<\/p>\n
The model underwent extensive pre-deployment testing for capabilities and safeguard calibration, along with post-deployment and third-party testing.<\/p>\n
Grok 4.6 is available today through Cursor and Grok Build, as well as via the API and partner platforms including OpenRouter, Vercel and Cloudflare.<\/p>\n
On the Artificial Analysis Intelligence Index, Grok 4.6 scored 61 points, matching GPT-5.6 Sol at its highest reasoning level and falling just one point short of Fable 5 Max. SpaceXAI says the model also outperforms Grok 4.5 across all listed benchmarks, including CursorBench, FrontierCode, APEX-Agents and Terminal-Bench.<\/p>\n
The company attributes the gains to a longer supplemental training run, stronger engineering data and expanded reinforcement learning focused on coding and knowledge work. Grok 4.6 is now 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, while a faster version costs twice as much. Cursor and Grok Build users will also receive twice their usual Grok 4.6 usage allowance during its first week.<\/p>\n","protected":false},"excerpt":{"rendered":"SpaceX AI has unveiled Grok 4.6, its latest model aimed at advanced coding, agentic tasks and knowledge work,…\n","protected":false},"author":2,"featured_media":137871,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10],"tags":[68122,6364,59562,68121,29448,2899],"class_list":["post-137870","post","type-post","status-publish","format-standard","has-post-thumbnail","category-xai","tag-chat-gpt-sol","tag-grok","tag-grok-4-6","tag-space-x-ai","tag-space-x","tag-xai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/137870","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=137870"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/137870\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/137871"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=137870"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=137870"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=137870"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}