August van Caleb

Master and Bachelor Degree from Zhejiang University.
AI Algorithm Engineer.
Embodied Intelligence.

[August van Caleb @AugustvanCaleb]

Time-Series-Master is available now!

No coding required. You can do your own deep learning or machine learning based time series prediction with just clicks!

Support LSTM, RNN, Transformer, XGboost, LightGBM, ARIMA etc.

You can also custom your own model.

Check it now!

[Arena.ai @arena]

DeepSeek-V4-Flash-20260731 (High) by @deepseek_ai landed in Agent Arena at #21 overall (+1.98% net-improvement), and among open source #3 based on +12.5K real-world agentic sessions.

This release is ranked 6 higher than DeepSeek-V4-Pro (-0.47%) and 13 higher than the previous DeepSeek-V4-Flash model (-2.81%).

Across key signals, DeepSeek-V4-Flash-20260731 (High) is strong in Confirmed Success (+6.99%), but weaker in Steerability (-2.31%).

Below we break down how it scored across all 5 signals, drawn from tasks submitted by a global community of users.

Here’s an overview on the signals:
- Confirmed Success: an explicit "yes that worked" feedback from the user
- Praise vs. Complaint: implicit sentiment in users reactions
- Steerability: can the model course-correct when you push back?
- Bash Recovery: how it recovers from CLI errors (primary signal for tool use)
- Tool Hallucination: does it call tools that don't exist

[August van Caleb @AugustvanCaleb]

@ollama well, your model inference speed is visually experiencing a dramatic slow. i used to be happy to enjoy one of the best token plan including glm 5.2 in your platform, maybe it's time to turn back to deepseek official api.

[August van Caleb @AugustvanCaleb]

Qwen3.8-max 太垃圾了

今天开了一个节省计划套餐，简单地做了几个UI，相较deepseek暴露四大问题：

1. 用户意图解析错误。一方面，对于用户输入中潜藏的含义解析不清。另一方面，在遇到用户输入实在较为含糊时，要简单地根据语境向用户确认。
2. 不知道清理设计样板文件
3. 容易搞坏原有代码设计
4. 模型推理速度慢。阿里的算力这么紧张吗？对比deepseek慢得要死，目测慢了一倍。

阿里这股价得使劲跌了

[Qwen @Alibaba_Qwen]

📢Meet Qwen3.8-Max — our most capable model to date.

Next week, the open weights of Qwen3.8-Max will be released, and Qwen3.8-27B is also going open-weights to meet you all!🎉

Qwen3.8-Max, a new bar for coding and cowork at 2.4T parameters:

- Autonomous coding: 10+ days of self-evolving development, from empty folder to production without hand-holding, complete project trace in the GitHub：https://t.co/iVHZWQoeSo
- Real work, real results: Production-quality deliverables across hundreds of professions.
- Long-horizon mastery: System-level autonomous planning with closed-loop adaptive learning, driving 500+ turns of chip design optimization and 365 days of e-commerce strategy.
- Native multimodal intelligence: Vision isn't just input — it's a continuous feedback loop for planning, execution, and self-correction.

💰Pricing:
Input: $2.0 / M tokens
Output: $6.0 / M tokens
Implicit Caching: $0.25 / M tokens

Start building with Qwen3.8-Max! 🚀

📖 Blog: https://t.co/iwjmQxLBof
✅ Qwen Studio: https://t.co/4V2pFvDovG
⚡ API: https://t.co/gAGqaLQGbN

[August van Caleb @AugustvanCaleb]

@_FORAB nice

[August van Caleb @AugustvanCaleb]

@yinsuecci 哈哈哈哈哈哈

[August van Caleb @AugustvanCaleb]

@keyahayek @grok 出国留学问新东方做规划，都会说哥大绝对保底。

新东方的政策是，如果都没录取上，全额退款。我猜测新东方靠哥大赚了不少钱。

[Alex Nguyen @alexcooldev]

After getting tons of feedback on the 3D chess game I vibe-coded in 2 hours, I just shipped another update:

→ 3 new battle systems
→ Two-player mode
→ Walking animations for the chess pieces
→ A complete sound system
→ Improved performance

And the crazy part is I only needed another hour of vibe coding to build all of this.

You can officially replace the default chess game on your MacBook now lol. 😆

[August van Caleb @AugustvanCaleb]

@BitcoinIntelX BTC

[August van Caleb @AugustvanCaleb]

@DRbitcoin36 这个年龄段的大学生像这样的太多了，但大部分连工作都找不到，结果她就凭一个标配直接跃龙门了

[August van Caleb @AugustvanCaleb]

@Fei2411 LSTM能出来不是没道理的。长程确定性的记忆是很重要的。

[August van Caleb @AugustvanCaleb]

agree, though we may still have a long way to go. the current problem is about tools not AI itself.

[Elon Musk @elonmusk]

This is exactly right. Source code is on the verge of becoming like assembly.

The next step is getting rid of “source code” entirely and just making an efficient binary directly with AI.

[August van Caleb @AugustvanCaleb]

@xie_Dl 画风都不一样

[August van Caleb @AugustvanCaleb]

@Michaelzsguo 社会背景不一样，思考的东西也不一样

[August van Caleb @AugustvanCaleb]

@jinchenma_ai 感觉不是个好方向

[August van Caleb @AugustvanCaleb]

@ValaAfshar Happy to see you guys are so obsessed with Chinese stuffs.

[August van Caleb @AugustvanCaleb]

详细计算可参照
https://t.co/dkrGc49oCp

图片里则是我现在查到的GLM官方价格

[August van Caleb @AugustvanCaleb]

计算了一下ollama cloud的glm 5.2的实际额度。

这周从周一到周五，一共用掉如图115.6%的额度。
由于都是通过pi来用，pi自身有token计数功能，统计得到总共token用量为=25,916,973+211,008,278+16,950,276=253,875,527，
则100%额度对应219,615,507.79的glm5.2的token。

因此ollama cloud的100%本周对应约2.2亿token。

更详细的：
输入大约252.37M，输出大约1.51M。拿智谱官方API价格来算的话，以90%都是cache来估计，相当于一周用掉了698元。

则按月来算，不考虑kv cache，以官方API计价，ollama cloud的价格相当于7600元；考虑KV Cache，则相当于2500-3500元。

[Santiago @svpino]

My transition away from Anthropic is now complete.

I love their models. I love Claude Code. But I'm not willing to keep sending them money while they keep lobbying the government to stifle competitors.

I hope Anthropic finds its way, and I'd be happy to consider them again when they do.

[Max For AI @MaxForAI]

API接口的规范居然会影响模型的智能和成本消耗🤯？？

刚刚， @OpenAI 公布了一个很反常识的结果：

同一个GPT-5.6 Sol，没有重新训练，也没有更换模型，只调整了API调用方式，ARC-AGI-3公开集得分就从13.3%提升到38.3%，接近原来的3倍

但输出Token反而减少到原来的六分之一。

原因是OpenAI发现，ARC原本的测试框架会在每次操作后丢弃模型的私有推理，并在上下文过长时直接删除早期记录。

模型虽然刚刚摸索出游戏规律，下一轮却又要从头理解，像是一边做题，一边被持续清空记忆。

随后，OpenAI改用Responses API，开启retained reasoning和compaction，让模型能够跨轮保留推理状态，并在上下文过长时压缩历史，而不是直接截断。

结果模型思考得更少，策略更连贯，能力和效率同时大幅提升。

终于有人发现API已经不只是调用模型的接口，它开始直接决定模型表现出来的智能，以及完成任务需要消耗多少Token。

相同的模型权重，因为状态管理和上下文处理方式不同，最终可能表现得像两个代际的模型。

以后评测一个模型，可能不能只写模型名称了。

还得写清楚用了什么API、是否保留推理、上下文如何压缩。

否则测出来的未必是模型有多聪明，也可能只是调用框架有多蠢。

之前国内模型API这个问题一直都存在，各家API都乱的不行😅。。。