AnswersPublished23 sources
Grok 4.7: xAI’s New Model for Coding and Knowledge Work
Grok 4.7 launched on September 21, 2026 as xAI’s new model for coding, agentic tasks and knowledge work. The model has a 500,000 token context window and is available as grok 4.7 through the xAI API, with integrations documented for Grok Build and Cursor.
Published byEdited with GPT-5.6 TerraImages generated with GPT Image 2
Research answer
Key takeaways
- Grok 4.7 launched on September 21, 2026 as xAI’s new model for coding, agentic tasks and knowledge work.
- The model has a 500,000 token context window and is available as grok 4.7 through the xAI API, with integrations documented for Grok Build and Cursor.
- The reported September 12 launch target slipped: Grok 4.7 was announced on September 21.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is xAI’s newly released Grok 4.7—launched after delays from Elon Musk’s original September 12 target—including its coding and knowledge. Article summary: Grok 4.7 is xAI’s September 21, 2026 flagship for coding, agentic work, and professional knowledge work, positioned as a longer-running, more self-checking successor to Grok 4.6 at the same standard token price. [3] ## W. Topic tags: general, general web, user generated, documentation. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks,
Grok 4.7 is xAI’s latest flagship model for coding, agentic tasks and professional knowledge work. The key practical change is not a lower sticker price: it retains Grok 4.6’s standard API rates while xAI says it can work for longer, check its work more carefully and handle more difficult tasks. 25
32
What Grok 4.7 is designed to do
xAI positions Grok 4.7 around code, tool-using agents and knowledge-work workflows. The company says the model uses a larger base model than Grok 4.6 and received a longer reinforcement-learning run on a harder task mix. Its stated goal is better performance on longer-running work rather than just short, single-turn answers. 32
For developers, the documented model identifier is grok-4.7. It accepts text and image inputs, generates text output, and has a 500,000-token context window. 19
31
That makes the release most relevant to teams building coding assistants, agent loops, research or document workflows, and tool-connected applications—not simply chat experiences.
Grok 4.7 pricing
For API prompts below 200,000 tokens, xAI lists the following rates per million tokens: 25
| Token type | Price per 1 million tokens |
|---|---|
| Input | $2.00 |
| Cached input | $0.50 |
| Output | $6.00 |
For long-context prompts of 200,000 tokens or more, rates rise to $4.00 for input, $1.00 for cached input and $12.00 for output per million tokens. 25
The important comparison for existing Grok 4.6 users is that the headline $2 input / $6 output rate remains unchanged. A lower token price does not automatically mean a lower total task cost, however: long-running agents can use many more input and output tokens, so teams should measure cost on representative workloads.
Availability: API, Grok Build and Cursor
Grok 4.7 is directly available through xAI’s API. xAI’s documentation also shows it in Grok Build, the company’s coding-agent environment, and references Cursor integrations. 19
21
27
The API supports tool calling, including built-in tools such as web search, X Search, code interpretation and collections search, as well as developer-defined function calls. Those capabilities can be combined with Grok 4.7 in an application or agent workflow. 26
xAI also documents a US regional API endpoint that serves Grok 4.7 and Grok 4.6 for teams that need US request handling and model inference. 22
Reported benchmark results versus Grok 4.6 and GPT-5.6 Sol
xAI reported benchmark comparisons using Grok 4.7 at xhigh effort, Grok 4.6 at high effort, and GPT-5.6 Sol at max effort. These are useful directional results, but they are not a substitute for independent, like-for-like evaluations—especially because reasoning-effort settings differ. 34
| Benchmark | Grok 4.7 | Grok 4.6 | GPT-5.6 Sol Max |
|---|---|---|---|
| CursorBench 4.0 | 46.3% | 40.4% | 41.7% |
| DeepSWE v1.1 | 71.0%* | 65.2% | 72.7% |
| EEBench | 64.0% | 53.0% | 39.4% |
| AA Briefcase v1.1 | 1,657 | 1,546 | 1,487 |
| Terminal-Bench 4.0 | 38.0% | 20.3% | 37.3% |
| Harvey Legal Agent | 19.6% | 15.8% | 2.5% |
| HealthBench Professional | 56.7% | 48.5% | 60.5% |
*xAI reported the DeepSWE result at high effort rather than xhigh. 34
On those reported figures, Grok 4.7 improves on Grok 4.6 across the table and exceeds GPT-5.6 Sol on several listed evaluations. But it does not lead everywhere: it trails GPT-5.6 Sol on DeepSWE and HealthBench Professional. 34
43
Safety and cybersecurity claims
xAI says Grok 4.7 has a new safeguard stack and is its strongest tested model for refusal behavior and jailbreak resistance. The company reported a 62.4% score on LatchBio’s biosafety benchmark and said that 3.3% of risky dual-use prompts passed through on HackerBench v0.3, while legitimate security work was rarely blocked. 33
37
Those figures should be read as vendor-reported safety evidence, not a complete guarantee of safe behavior in deployment. Safety results can depend on the prompt set, system configuration, tool access and the definition of a harmful or legitimate request.
xAI also said selected cybersecurity partners would receive invite-only red-team access for defensive research. 33
Rollout timeline and unconfirmed claims
A reported September 2 statement from Elon Musk indicated a 10-day timeline, implying a September 12 target. That date passed without a release; xAI announced Grok 4.7 on September 21, 2026. 7
32
The supplied official launch materials confirm a larger base model and longer reinforcement-learning training, but they do not provide a parameter count or identify SpaceX engineering data as part of Grok 4.7’s training. Therefore, claims that the model has 2.1 trillion parameters or was trained on SpaceX data should be treated as unconfirmed based on the materials available here. 32
31
Likewise, the supplied official sources do not confirm plans for a future Grok 5 model to use SpaceX data. The defensible takeaway is narrower: Grok 4.7 is a released, 500,000-context model aimed at longer coding and knowledge-work tasks, available at the same standard API token rates as Grok 4.6. 19
25
32
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Grok 4.7: xAI’s New Model for Coding and Knowledge Work"?
Grok 4.7 launched on September 21, 2026 as xAI’s new model for coding, agentic tasks and knowledge work.
What are the key points to validate first?
Grok 4.7 launched on September 21, 2026 as xAI’s new model for coding, agentic tasks and knowledge work. The model has a 500,000 token context window and is available as grok 4.7 through the xAI API, with integrations documented for Grok Build and Cursor.
What should I do next in practice?
The reported September 12 launch target slipped: Grok 4.7 was announced on September 21.