SpaceXAI Unveils Grok 4.7: A Leap in AI Capabilities
SpaceXAI has announced that Grok 4.7 is engineered to tackle complex tasks for extended durations, enhance output verification, and adeptly manage longer contexts when compared to its predecessor, Grok 4.6.
Under the vision of Elon Musk, SpaceXAI has introduced Grok 4.7, its most advanced artificial intelligence (AI) model tailored for coding, autonomous tasks, and professional intellectual pursuits, as reported by ET Tech.
The organization asserts that Grok 4.7 has been meticulously designed to handle intricate activities over prolonged intervals while improving output verification and contextual understanding significantly over Grok 4.6.
The developmental process featured an extensive reinforcement-learning protocol concentrated on tasks that necessitate several hours for completion.
The model boasts an impressive 500,000-token context window and accommodates both text and image inputs. It is accessible via the Grok API, Grok Build, Cursor, and various other developer platforms.
Performance Overview of Grok 4.7
Benchmark analyses conducted by SpaceXAI reveal a mixed performance profile when juxtaposed with other leading models.
In the CursorBench 4.0 assessment, which scrutinizes software engineering capabilities, Grok 4.7 achieved a score of 46.3%, an improvement from Grok 4.6’s 40.4%, while GPT-5.6 Sol and Fable 5.1 scored 41.7% and 51.8%, respectively.
On the DeepSWE v1.1 test, Grok 4.7 garnered a score of 71%, surpassing Grok 4.6’s 65.2% yet falling short of GPT-5.6 Sol’s 72.7%. Fable 5.1 registered a score of 70% in the same evaluation.
Additionally, SpaceXAI reported improvements across various domains, including electrical engineering, legal expertise, and multi-hour office operations.
Nonetheless, Grok 4.7 lags behind some competing systems in several assessment benchmarks, particularly in terminal operations and clinical reasoning.
Emphasis on AI Safety
The Grok 4.7 model integrates a newly developed safety system aimed at bolstering its resilience against jailbreak attempts and managing potentially harmful requests.
SpaceXAI indicated that the model was crafted to strike a balance between facilitating legitimate cybersecurity tasks and imposing restrictions on high-risk usage.
In its HackerBench v0.3 test evaluating risky and malicious cyber requests, only 3.3% of dual-use prompts were permitted, while legitimate cybersecurity operations faced minimal blockage.
Furthermore, the company highlighted that Grok 4.7 excelled in LatchBio’s biosafety benchmark, achieving a commendable score of 62.4%.
Pricing and Availability of Grok 4.7
The cost structure for Grok 4.7 is set at $2 per million input tokens and $6 per million output tokens for standard usage. SpaceXAI offers a more rapid variant at double the token rates, ensuring enhanced output speeds.
This model is accessible through the Grok API, Cursor, and Grok Build platforms, along with third-party coding tools, model routers, and cloud services.
This broad accessibility empowers developers to integrate Grok 4.7 into their coding and autonomous workflows, rather than relying solely on chatbot interfaces.
Source link: Enterpriseai.economictimes.indiatimes.com.
