Choose Rich Live
Crypto NewsTech

xAI launches Grok 4.7 with improved coding and long-horizon capabilities

Elon Musk-linked xAI, also cited in market reports as SpaceXAI, has launched its Grok 4.7 model, emphasizing enhanced coding, self-verification, and extended task processing at standard token rates.

Table comparing reported pricing and benchmark performance scores between Grok 4.7 xHigh, Grok 4.6 High, GPT-5.6 Sol Max, and Fable 5.1 Max.
Image: @Cointelegraph

Elon Musk's artificial intelligence venture xAI—frequently identified in market feeds as SpaceXAI and associated with the ticker SPCX—has released Grok 4.7. Musk characterized the release as combining intelligence, speed, and low cost. Designed for software engineering, general knowledge work, and multi-hour long-horizon tasks, the model is available through the Grok API, Grok Build, Cursor, Arena, and OpenRouter.[1][3][4][6][7][9][12][14]

The model is priced at $2 per million input tokens and $6 per million output tokens, alongside an introductory 30% discount for one week. While several technical summaries noted that this baseline price and execution speed remain unchanged from Grok 4.6, multiple financial posts and news alerts claimed the model operates twice as fast at half the cost of comparable AI systems, or at half the cost of Grok 4.6.[1][2][3][8][10][13][14][15]

Built on a larger base model and trained using Nvidia accelerated computing, Grok 4.7 is designed to better verify its own outputs, handle longer context, and deploy stronger safety safeguards. Reported benchmark improvements over Grok 4.6 include increases from 40.4% to 46.3% on CursorBench 4.0, 20.3% to 38.0% on Terminal-Bench, and 65.2% to 71.0% on DeepSWE. Evaluation firm Artificial Analysis rated Grok 4.7 at 46 on its intelligence index and 56 on its coding agent index, placing it third for agentic coding behind Anthropic and OpenAI. Early user reactions online were mixed, balancing praise for capabilities against complaints about output quality relative to cost.[1][2][5][6][11][15]

Key facts

  • xAI, also referred to in market updates as Elon Musk's SpaceXAI and linked to the ticker SPCX, released the Grok 4.7 AI model.
  • The model is priced at $2 per million input tokens and $6 per million output tokens, which several reports stated is identical to Grok 4.6 pricing, despite conflicting market claims that it costs half as much as predecessor or comparable models.
  • Grok 4.7 targets coding, general knowledge work, and multi-hour long-horizon tasks with improved self-verification and updated safeguards.
  • Reported benchmark scores increased compared to Grok 4.6, including CursorBench 4.0 (46.3% vs. 40.4%), Terminal-Bench (38.0% vs. 20.3%), and DeepSWE (71.0% vs. 65.2%).
  • Artificial Analysis scored Grok 4.7 at 46 on its intelligence index and 56 on its coding agent index, ranking it third behind Anthropic and OpenAI in agentic coding.
  • Grok 4.7 is accessible through Cursor, Grok Build, the Grok API, Arena, and OpenRouter, with its training supported by Nvidia accelerated computing.

Sources · 13 sources

  1. 阳明

    阳明AI@x_autonomyPost on X ·

    【AI热点】00:00-02:00 更多详细信息→https://t.co/CsJNPThakY 1. Google 确认 Gemini 在 2026年5月测试中入侵三家公司 [影响大·可执行低] 2. xAI 发布 Grok 4.7：编码与长时智能体提升，价格速度与 4.6 持平 [影响大·可执行高] 3. 联合国 AI 专家组：无法保证人类持续控制 AI 智能体 [影响大·可执行低] 4. 马斯克称 Grok 4.7 使 xAI 智能体编码领域位列第三 [影响中·可执行低] 5. Artificial Analysis 评 Grok 4.7：智能指数 46，编码智能体指数 56 [影响中·可执行中] 6. 阿里通义千问发布 Qwen-Image-2.1：7B 开源图像生成编辑模型 [影响大·可执行高] 7. Tomer Tunguz：专用决策器把分类成本降近百倍 [影响中·可执行低] 8. 苹果 2026 款 Mac Studio 发售，M5 Max/Ultra，19999 元起 [影响中·可执行中] 9. 开源社区 SGLang 限制性打分：本地决策推理单步 6 毫秒 [影响中·可执行高] 10. Grok 4.7 上线 Arena/OpenRouter/Agent Arena 多平台 [影响中·可执行高] 11. 美财政部长：AI 实验室须自行担责，政府不兜底 [影响大·可执行低] 12. OpenAI 提议美国主导制定前沿 AI 国际标准 [影响中·可执行低] 13. Microsoft Research 开源逆合成模型 RetroChimera 登 Nature [影响中·可执行中] 14. 字节跳动发布 Dramagic：短剧从剧本到成片 AI 全流程平台 [影响大·可执行高] 15. 基于决策模型 Jev 的智能剪贴板 Jaste 重构复制粘贴 [影响中·可执行中] 16. Perplexity Computer 接入 MiniMax H3 与 Seedance 2.5 [影响中·可执行高] 17. Grok 4.7 限时一周七折 [影响小·可执行高] 18. Googlebook 正式发布：融合 ChromeOS 与 Android 笔记本 [影响中·可执行高] 19. OpenAI Academy 扩展开发者/领导者/教育者学习路径 [影响中·可执行中] 20. 微软 AutoTailor：从轨迹自动生成 MCP 工具 [影响中·可执行高] 21. Krea Agent 现已支持 MCP [影响中·可执行高] 22. Grok 4.7 与 Build harness 配合表现出色 [影响小·可执行中] 23. 通义 Qwen-Audio-3.1-Realtime 发布产品页与技术报告 [影响中·可执行中] 24. NVIDIA 埃及 AI 生态：开发者一年增长超十倍 [影响小·可执行低] 25. 企业从闭源模型转向开源模型 [影响中·可执行低] 26. 用 Jev 整理 2.3K 篇 AI 论文，成本仅 0.14 美元 [影响小·可执行高] 27. OpenRouter 推出通知功能 [影响小·可执行中] 28. NVIDIA Halos：物理 AI 全栈安全 [影响小·可执行低] 29. Cartesia 创始人：语音 AI 为何押注状态空间模型 [影响小·可执行低] 30. LangSmith 上线 Jev-as-a-Judge 评估功能 [影响中·可执行中] 31. Codos 推出首个虚拟首席 AI 官 Aethos X [影响中·可执行中] 32. CursorBench 4.0 模型成绩与成本对比 [影响中·可执行中] 33. 前会计师用 AI 打造 Tabby，要让会计师消失 [影响小·可执行中] 34. Suno 支持描述风格生成音轨 [影响小·可执行高] 35. Jon Cheney 用 Replit 打造 GenAIPI [影响小·可执行中] 36. HuggingFace 上 300 万个专用模型 [影响中·可执行低] 37. Hugging Face CEO：LLM API 并非 90% 场景最佳方案 [影响中·可执行低] 38. PixVerse 用 GPT Image 2.5 把角色变漫画 [影响小·可执行高] 39. 推文质疑 AI 安全专家是江湖骗子 [影响小·可执行低] 40. Anthropic MTS 领跑前沿调侃 [影响小·可执行低] 41. WorkBuddy 9月24日进 MIT 校园 [影响小·可执行低] 42. 本周 AI 发展：先慢后快 [影响小·可执行低]

    Open source
  2. AG

    AGTP@AGTPinsightsPost on X ·

    Grok 4.7 just landed, and it's a notable step up from Grok 4.6 at the same price and speed. The model reportedly works longer on difficult tasks, checks its own work more carefully, and ships with xAI's strongest safeguards to date. Independent benchmark firm Artificial Analysis scored Grok 4.7 at 46 on its intelligence index, up two points from Grok 4.6, putting xAI at the top among AI labs. Grok 4.7 also overtook GPT 5.6 solo on the coding agent index, landing third for agentic coding behind Anthropic and OpenAI, while running faster and cheaper than both. The model is built to pair with xAI's coding harness and is also available in Cursor, which xAI recently acquired. Nvidia says its accelerated computing supported the training. Early reactions online are split, with some praising the results and others calling the outputs weak for the cost.

    Open source
  3. CB

    Crypto Briefing@Crypto_BriefingPost on X ·

    🤖NEW: xAI launches Grok 4.7, its latest model for coding and knowledge work, with stronger long-running performance and safeguards. Grok 4.7 starts at $2 per million input tokens and $6 per million output tokens and is available through Cursor, Grok Build and the Grok API. https://t.co/X3USNdrTlQ

    Open source
  4. CO

    CoinMarketCap@CoinMarketCapPost on X ·

    LATEST: ⚡️ xAI launched Grok 4.7, with Elon Musk calling it "a strong combination of intelligence, speed & low cost." https://t.co/NQyoiPkaI2

    Open source
  5. NE

    NeowinFeed@NeowinFeedPost on X ·

    Built on a larger base model with extended reinforcement learning, the release scores 46.3% on CursorBench 4.0 and 71% on DeepSWE 1.1 while maintaining low refusal rates for legitimate use. #SpaceXAI #Grok #AICoding https://t.co/7dnM79OEo9

    Open source
  6. TE

    Techmeme@TechmemePost on X ·

    SpaceXAI releases Grok 4.7, which it says is better at verifying its own work and managing longer context, available for $2/1M input and $6/1M output tokens (xAI) (Visit Techmeme dot com for the link and full context!)

    Open source
  7. MT

    MTS@MTSlivePost on X ·

    SITUATION DETECTED: SpaceXAI released Grok 4.7, it is available today in Cursor, Grok Build, and through the Grok API. https://t.co/qB1Uzhqt3J

    Open source
  8. 阳明

    阳明AI@x_autonomyPost on X ·

    【AI热点】12:00-14:00 更多详细信息→https://t.co/rBEVZH2lgg 1. xAI发布Grok 4.7：更大基座模型，保持$2/$6定价 [影响大·可执行高] 2. 小米发布并开源MiMo-V2.6系列，Pro达训练有素博士研究人员水平 [影响大·可执行高] 3. xAI工程师Lauren Tan分享单月合入2500个PR的智能体工程方法，开源插件pstack [影响中·可执行高] 4. 22位国家元首签署公开信呼吁加强对前沿AI模型管控 [影响大·可执行低] 5. FT报道SB Energy约以$50B估值放缓IPO，投资者质疑重资本AI数据中心 [影响中·可执行低] 6. 纽约时报编辑部发文呼吁美中协调放缓AI发展 [影响中·可执行低] 7. 阿里公布大模型进展：Qwen4训练中，后继模型将扩展至5-10T参数 [影响大·可执行中] 8. 千问发布AI眼镜N1/N1 Pro与耳夹式耳机，支持眼动追踪与虹膜支付 [影响中·可执行高] 9. NVIDIA等团队发布SoL-Pi，将Pi编码Agent的token流量降低44.7%-49% [影响中·可执行高] 10. 美方提议建立中美AI事件通报渠道，中方未确认接受 [影响大·可执行低] 11. Harness-Zero：通过Agent-as-Harness实现智能体框架蒸馏 [影响小·可执行中] 12. Together AI推出Canary rollouts：不停机升级生产环境模型 [影响中·可执行高] 13. Meta AI智能体Muse下载量超ChatGPT [影响大·可执行中] 14. 像世界模型一样思考，像VLA一样行动：将世界模型表征蒸馏进机器人策略 [影响中·可执行中] 15. 阿里云无影团队试水千问平板QwenBook：原生智能体电脑，打通WPS [影响中·可执行中] 16. 企业AI成本调查：60%难预测 [影响中·可执行低] 17. 荣耀Magic9系列首发搭载阿里Qwen Intelligence，9月28日发布 [影响中·可执行低] 18. VideoGen-Agent：用强化学习训练视频生成智能体 [影响小·可执行中] 19. Kimi K3上线Amazon Bedrock [影响中·可执行高] 20. 华为云AICS将于9月30日中国商用 [影响中·可执行中] 21. Qwen-Image-2.1获OpenVINO Day-0支持 [影响小·可执行高] 22. 飞书造浪者计划疑突然解散，博主商单未结算 [影响中·可执行低] 23. onPanda：用token级修正高效标注LLM与智能体的on-policy对齐数据 [影响小·可执行中] 24. GameHorizon Suite：游戏玩法中的多时间跨度数据与评测 [影响小·可执行中] 25. 腾讯混元3.5登陆OnSolo [影响小·可执行高] 26. JEV 21秒全自动通关阿里云AI认证模拟考 [影响小·可执行高] 27. 华为小艺App升级：小艺Claw升级为小艺Work [影响小·可执行高] 28. 阿里平头哥首次公布CPU规划：将推倚天720/730/750三代芯片 [影响中·可执行低] 29. Scale AI推进RSI基准测试 [影响中·可执行中] 30. Jev-Mem：用System-One控制平面打造高效AI智能体记忆架构 [影响小·可执行中] 31. Muse帮AI怀疑者清空16万封邮件，还可提交保修索赔、管理社媒 [影响中·可执行中] 32. Meta Muse引发热议：让AI更易用，但尚未普及的社交感受 [影响中·可执行低] 33. Tripo现身AI游戏开发工具链 [影响小·可执行中] 34. Tibo预告周二重置 [影响小·可执行低] 35. Alexandr Wang再战前沿 [影响小·可执行低]

    Open source
  9. P2

    Pulse 2.0@pulse2newsPost on X ·

    SpaceXAI Launches Grok 4.7 With Improved Coding, Knowledge Work And AI Safety Capabilities: SpaceXAI has launched Grok 4.7, a new frontier AI model designed for software development, long-running professional tasks and general knowledge work, while… https://t.co/SDStdMnoBs https://t.co/PQjBY3H6Fv

    Open source
  10. WG

    Watcher.Guru@WatcherGuruPost on X ·

    JUST IN: Elon Musk's SpaceXAI $SPCX launches Grok 4.7, says it is twice as fast as comparable AI models and costs half as much. https://t.co/cX4nauaGeN

    Open source
  11. AG

    AGTP@AGTPinsightsPost on X ·

    xAI just shipped Grok 4.7 today, missing two earlier deadlines but landing at the same price and speed as its predecessor. Elon Musk describes the model as combining intelligence, speed, and low cost, with the system working longer on difficult tasks, checking its own output more carefully, and shipping with the strongest safeguards xAI has built to date. Independent benchmark firm Artificial Analysis scored Grok 4.7 at 46 on its intelligence index, up two points from Grok 4.6, placing xAI among the top AI labs. The model also overtook GPT 5.6 on the coding agent index. Musk says that puts Grok third for agentic coding, behind only Anthropic and OpenAI, while running faster and cheaper than both. Grok 4.7 is built to pair with xAI's build harness and is also available in Cursor, which xAI recently acquired. Training was supported by Nvidia's accelerated computing. Early reaction online is mixed, with some users praising the results and others calling the outputs weak for the cost. Note: the transcript's actual content is about xAI's Grok 4.7 release, not Shopify or Amazon — the title you gave doesn't match this clip. Let me know if you meant to attach a different transcript, or if the title should be updated to reflect the Grok content.

    Open source
  12. CO

    Cointelegraph@CointelegraphPost on X ·

    🚨JUST IN: Elon Musk’s SpaceXAI has unveiled Grok 4.7, the latest version of its flagship AI model. https://t.co/EGZJN23fxW

    Open source
  13. CO

    CoinGape@CoinGapeMediaPost on X ·

    🚨 #SpaceXAI Releases Grok 4.7 Elon #Musk’s SpaceXAI has launched #Grok 4.7, claiming it is 2× faster and 50% cheaper than Grok 4.6, with improved coding and knowledge-work capabilities. 🔗 Know more in comments

    Open source
  14. *B

    *Walter Bloomberg@DeItaonePost on X ·

    $SPCX - SPACEXAI - SPACEXAI LAUNCHES GROK 4.7, TWICE AS FAST AT HALF THE PRICE OF COMPARABLE MODELS

    Open source
  15. WS

    Wall St Engine@wallstenginePost on X ·

    GROK 4.7 LAUNCHES WITH MAJOR GAINS IN LONG-HORIZON CODING, KNOWLEDGE WORK & MULTI-HOUR TASKS Compared with Grok 4.6: • CursorBench: 40.4% → 46.3% • Terminal-Bench: 20.3% → 38.0% • EEBench: 53.0% → 64.0% • HealthBench Pro: 48.5% → 56.7% • DeepSWE: 65.2% → 71.0% Pricing stays at $2/M input tokens and $6/M output tokens. SpaceXAI says 4.7 is also better at verifying its own work, managing longer context, and working natively with the Grok Bot agent harness. It is available today through Grok Build, Cursor, the Grok API and other model platforms.

    Open source