xAI Launches Grok 4.7 for Extended Knowledge Work
On September 21, 2026, SpaceXAI released Grok 4.7, its newest frontier model tailored specifically for coding and professional knowledge work. Rather than raising prices for upgraded capabilities, xAI has kept Grok 4.7 at the same cost and inference speed as its predecessor, Grok 4.6. The model is priced at $2 per million input tokens and $6 per million output tokens, though a fast variant is available that offers twice the output speed for double the price.
The primary architectural shift with this release is endurance. Grok 4.7 is built on a new, larger base model that underwent an extended reinforcement learning run. xAI specifically weighted this training toward difficult, multi-hour tasks, aiming to improve the model’s ability to manage long contexts and verify its own work before returning an output.
For developers and operators building around the Grok API, the model was also trained to natively understand the Grok Bot harness. According to xAI, this makes the model better at conversational tasks, general knowledge work, and creating presentations and documents.
Benchmark Performance and Security Improvements
xAI tested Grok 4.7 against a suite of benchmarks designed to stress-test professional and technical execution, comparing it against models like Fable 5.1, GPT-5.6 Sol, Sonnet 5, and GPT-6 Astra.
On DeepSWE v1.1, a software engineering benchmark, Grok 4.7 posted a high-effort score of 71.0%, edging out Fable 5.1's 70.0% and improving upon Grok 4.6's 65.2%. It also reached 46.3% on CursorBench 4.0, which stresses longer-running coding tasks, and 38.0% on Terminal-Bench 4.0 for multi-hour terminal work.
The company highlighted the model's performance on tasks typically done by professionals like lawyers, financial analysts, and nurses. Grok 4.7 scored 1695 on the GDPval Elo ranking, sitting between GPT-6 Astra (1542) and Fable 5.1 (1735). It reached 1,657 on the AA Briefcase v1.1 benchmark for multi-hour office work. In more specialized domains, Grok 4.7 posted 56.7% on HealthBench Professional, 64.0% on EEBench for electrical engineering, and 19.6% on the Harvey Legal Agent Benchmark.
Alongside performance upgrades, xAI deployed an entirely new safeguard stack. In dual-use domains like cybersecurity and biology, Grok 4.7 scored 62.4% on LatchBio’s biosafety benchmark. On xAI's own HackerBench v0.3, the model allowed only 3.3% of risky, malicious cyber prompts to pass through, while maintaining low refusal rates for legitimate security tasks. xAI is also granting select partners invite-only access to the model’s red-team capabilities for defense research.
What Grok 4.7 Means for SMB Automation
For small and mid-sized businesses building automated operations, an AI model's ability to stay on track during long-running tasks is a practical necessity. Automations often break when a model loses context mid-process or fails to check intermediate steps before passing flawed data to the next tool in the chain.
Grok 4.7’s focus on multi-hour task endurance and self-verification directly addresses this reliability gap. If a model is actively trained to pause and verify its own work over longer contexts, it reduces the amount of human intervention required to babysit background workflows. Practically, this means more stable automations for parsing complex vendor contracts, reconciling monthly financial reports across multiple documents, or auditing internal databases.
Pricing is the other major operational takeaway. By delivering a larger base model and better performance at the exact same $2/$6 price point as Grok 4.6, xAI is keeping the entry cost low for high-end professional knowledge work. SMBs running high-volume automations—such as sorting hundreds of daily customer support tickets or processing heavy legal and financial documents—can slot Grok 4.7 into their existing pipelines without recalculating their API budgets.
The addition of native Grok Bot harness understanding and better document creation also simplifies the final mile of automation. When a model can generate a clean, formatted brief or presentation at the end of a long research task, operators spend less time stitching together separate formatting scripts. Ultimately, Grok 4.7 offers a highly competitive, economically predictable engine for businesses relying on autonomous, multi-step operations.
Grok 4.7 is available today in Cursor, Grok Build, the Grok API, and through third-party coding harnesses and cloud platforms.