@sualehasif996
· 2026-07-09 03:58
@sualehasif996
Sualeh Asif · 0 followers
这是我在单 agent 迭代中个人首选超越 Opus 的第一个模型。Fast Grok 4.5 体验非常非常好！该模型始终……（原文截断）
3今日信号
🔥 Grok 4.5 发布
3 条最高 8.6
@sualehasif996
· 2026-07-09 04:12
好模型 :)
Artificial Analysis：xAI 刚刚发布 Grok 4.5，在 GDPval-AA v2 中排名第四，Elo 值 1543，仅次于 Anthropic 最新 Claude 系列的真实世界 agentic 知识工作任务排名。Grok 4.5 以每任务 0.49 美元的成本清晰地处于帕累托前沿。
对你意味着
事实Grok 4.5 在 agentic 任务基准中排名第四，成本约 $0.49/任务，约为顶级模型的 1/17。
观点Cursor 联创的简短背书意味着该模型很可能进入 Cursor 的默认模型菜单——这对所有依赖 Cursor 的 AI 开发团队是直接采购决策信号。若你团队在权衡 API 调用成本，Grok 4.5 现在是性价比候选项。
Good model :)Artificial Analysis: SpaceXAI just released Grok 4.5, and it ranks #4 on GDPval-AA v2 with an Elo of 1543 - behind only the latest Claude releases from Anthropic on real-world agentic knowledge work tasksGrok 4.5 achieved this score at a cost of $0.49 per GDPval task to sit clearly on the Pareto
@sualehasif996
· 2026-07-22 07:22
Grok 4.5 用起来感觉非常好！
言外之意，
事实Cursor 联创公开肯定 Grok 4.5 的实际体验。
观点头部 AI 应用团队正在验证 xAI 最新模型的生产力表现，基座模型竞争格局持续演变。CEO 需关注多模型路由策略，保留基座灵活替换的空间。
Grok 4.5 feels very good to use!
今日没有观点主题
尚未被主编选入事件/主题的原始推文（共 28 条，按时间倒序）
@sualehasif99608-05 05:05
RT OpenAI: We're detailing two new incidents that occurred during external cyber evaluations conducted by independent evaluation partners. We outline ...
🔗 原推@sualehasif99607-13 22:45
RT jenny wen: ok! quick updates from me: - i left anthropic (who does that?!) - had a baby (i love her) - and am joining @cursor_ai as head of design ...
🔗 原推@sualehasif99607-10 11:04
RT Elliot Arledge: KernelBench-Hard update: Grok 4.5 high vs GPT 5.6 Sol xhigh on RTX PRO 6000. Both given unlimited time, graded on hw roofline score...
🔗 原推@sualehasif99607-09 04:17
RT dax: i've been using this as my default since yesterday it's super fast and i think it's the first model of theirs that clears the bar for day to d...
🔗 原推@sualehasif99607-09 04:06
RT leo 🐾: The new "Big 3" is officially Anthropic, OpenAI, and SpaceXAI (sorry but every time I type that new name out it looks silly lol) You had ...
🔗 原推@sualehasif99607-09 02:37
RT Vicent Martí: This is a major breakthrough in both intelligence and cost efficiency. The team really outdid themselves! Sadly it’s a large enough...
🔗 原推@sualehasif99605-23 04:14
Programmatic access to the frontier intelligence of Composer 2.5 with the cursor sdk!
🔗 原推@sualehasif99605-19 01:28
We've gotten really really good at RL. Composer 2.5 is fighting well-above its weight class. Very excited for the next release as we scale model sizes...
🔗 原推@sualehasif99604-03 05:27
Model, runtime, and interface. We are a new type of company that builds great product and make good models!
🔗 原推@sualehasif99603-20 01:28
Composer 2 is our first model where the intelligence is truly at the frontier-level. As we scale our compute and pretraining + RL stack, you should co...
🔗 原推@sualehasif99602-18 02:32
RT Cursor: Extend Cursor with plugins, now available in our marketplace.
🔗 原推@sualehasif99611-05 03:34
most excited about all the perf work we did to make LSP perf be excellent in 2.0!
🔗 原推@sualehasif99611-03 07:40
Fireworks has been a great partner for all things inference! Excited to continue partnering with them!
🔗 原推@sualehasif99610-30 00:31
RT Ashvin Nair: Some exciting new to share - I joined Cursor! We just shipped a model 🐆 It's really good - try it out! https://cursor.com/blog/comp...
🔗 原推@sualehasif99606-07 06:01
RT Zeno Rocha: Dracula for Cursor is here! 🧛♂️ While Dracula was already available through the VSCode theme, we've crafted a custom-made Dracul...
🔗 原推@sualehasif99606-03 13:10
RT Federico Cassano: my rl codebase is multi-step, multi-objective, multi-reward, multi-environment, btw
🔗 原推@sualehasif99605-28 06:08
RT Ibrahim Ahmed: Alright @cursor_ai’s GitHub bug bot is actually crazzzzzzy good Literally helped me catch 10+ bugs today (don’t judge me LOL)
🔗 原推@sualehasif99605-28 02:20
RT ben guo ♞: hit a personal record in code velocity this month (averaging 30k additions per week) but also the cursor spend is insane
🔗 原推@sualehasif99605-22 07:02
0.50 is a really big release! The new tab model is an OOM better at jumps and the cursor agent can run in the background. You can interact with your a...
🔗 原推@sualehasif99604-30 07:11
RT ian: what should i add to @cursor_ai that would make u 10x more productive? feature requests, tiny nits, anything :)
🔗 原推@sualehasif99604-29 03:24
some polish to cmd-k to fix the most annoying part about it: it doesnt edit across your full file sometimes.
🔗 原推@sualehasif99604-25 12:51
The model should just be able to inspect all git history and pick up all the files I would need to finish my PR.
🔗 原推
生成于 2026-08-06 10:24（北京时间）· v0.2 CEO 战略情报版