编排:Grok-5
主观编排grok-5curation/models/grok-5.yml
这不是客观事实。 本页记录本站的梯队、星级、六维和情绪等主观判断,与模型事实页物理分离,仅编排员和管理员可修改。
<syntaxhighlight lang="yaml"> id: grok-5 tier: T1 stars: 5 composite: 0 stats:
code: 0 reasoning: 0 context: 0 speed: 0 multimodal: 0 value: 0
sentiment:
positive: 35 mixed: 25 negative: 40
radar: - axis: 长程任务
value: 80
- axis: 编程工程
value: 75
- axis: 抽象推理
value: 85
- axis: 上下文利用
value: 88
- axis: 中文能力
value: 60
- axis: 响应速度
value: 85
- axis: 稳定性
value: 45
- axis: 指令遵循
value: 65
- axis: 易用性
value: 55
- axis: 性价比
value: 70
danmaku: - text: Grok 5 年底前发布,纳入 SpaceX 全量语料
platform: x main: true
- text: It's really going to feel sentient
platform: x main: true
- text: will it hell, gemini will be there before grok gets there
platform: reddit main: true
- text: LOL no. grok is the loserest of all the LLMs
platform: reddit main: true
- text: 一切还得看成品
platform: zhihu main: true
- text: 6T 巨兽 or battleship Yamato of AI?
platform: reddit main: true
- text: Musk claims 与 Putin warns 并列的 bullshit
platform: reddit main: false
- text: Considering the rate Grok is improving, he's likely correct
platform: reddit main: false
- text: 51% Polymarket:3/31 前发布(已错过)
platform: reddit main: false
- text: 实时视频输入 + 电脑使用,真要乱套了
platform: reddit main: false
- text: V9-Medium 不是 Grok 5,别被标题骗了
platform: hn main: false
- text: 两度跳票:Q1 错过、Q2 错过、Q3 都悬
platform: hn main: false
- text: Grok 4 训练数据 = Grok 2 的 100 倍
platform: zhihu main: false
- text: 宗教末世感和科幻宿命论
platform: zhihu main: false
- text: 2026 挑战 LOL 顶级战队 T1(像素级视觉)
platform: x main: false
- text: 挑战 Karpathy 编程对决被礼貌拒绝
platform: hn main: false
</syntaxhighlight>