Kun

Models · Products · News · Insights
Building, testing & exploring AI.

[Kun @Kunuid]

@BeethKZ I claimed mine about 50 minutes ago, but the tokens might be all gone by now.
This is already the 3rd time ZCode has run a promotion like this, and there’s a solid chance there’ll be another drop next week.

[Kun @Kunuid]

Huge weekend drop from ZCode:
300 MILLION FREE tokens for GLM-5.3-Flash — the native multimodal model with 1M context window, on par with Claude Opus 4.8 at a fraction of the cost. Available worldwide 🌍

How to claim (international users):

1:Update ZCode desktop to v3.10.0+
2:Sign in with your https://t.co/4ylJZoFmql account (not BigModel)
3:Click the event banner in bottom-left corner after 15:00 Beijing time (07:00 UTC) Friday
4:Tokens activate at 20:00 Beijing time — switch to "Experience Plan" in model settings

⏳ Expires Monday 09:00 Beijing time. One per account, while supplies last.
@Zai_org @ZixuanLi_
#ZCode #GLM53Flash #AIProgramming #DevTools #FreeAI

[Kun @Kunuid]

Confirmed it works — I’m using it right now, go check it out.

Just update ZCode to v3.10.0 or above, then sign in with your https://t.co/4ylJZoFmql account. If you don’t have one yet, you can log in directly with Google.

Once you’re signed in, a pop-up for the 300M token giveaway will appear in the bottom-left corner of the ZCode UI.

[Kun @Kunuid]

@blankspeaker So when are they releasing the Android app for Grok BOT?

[Kun @Kunuid]

@ZixuanLi_ 你牛大了

[Kun @Kunuid]

@XiaopuPeng @Zai_org 不是，哥们，这下真牛来了。
你牛大了。

[Kun @Kunuid]

GLM just killed the game. Unmatched price-performance.

1M context. Natively multimodal. MIT open source.

GLM-5.3-Flash is live today.
Official list price: Input $0.15 | Cached $0.03 | Output $0.50

Vs current flagships (same basis, per 1M tokens):

OpenAI GPT-5.6 Sol
$4.00 / $0.40 / $20.00
27× input, 13× cache, 40× output

Anthropic Claude Opus 5
$5.00 / $0.50 / $25.00
33× input, 17× cache, 50× output

Google Gemini 3.1 Pro
$2.00 / $0.20 / $12.00
13× input, 7× cache, 24× output

xAI Grok 4.6
$2.00 / $0.50 / $6.00
13× input, 17× cache, 12× output

Meta Muse Spark 1.2
$1.25 / — / $4.25
8× input, 8.5× output

DeepSeek V4-Pro
$1.32 / $0.044 / $3.96
9× input, 1.5× cache, 8× output

Alibaba Qwen3.8 Max
$2.00 / $0.25 / $6.00
13× input, 8× cache, 12× output

Moonshot Kimi K3
$3.00 / $0.30 / $15.00
20× input, 10× cache, 30× output

50% off through Sep 9: Input $0.075 | Cached $0.015 | Output $0.25.

[Z.ai @Zai_org]

Introducing GLM-5.3-Flash

- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips

Blog: https://t.co/tzOmB7gdZP

Available now across all official platforms:

Weights: https://t.co/9LRMahY9Wa
API: https://t.co/VcaQnzYmS9
Coding Plan: https://t.co/Nk8Y98HNhU
ZCode: https://t.co/Peepqv4XSx
Chat: https://t.co/WCqWT0qCQb
AutoClaw: https://t.co/aGEG5HqTTb

[Kun @Kunuid]

@bpjzy @superwhisper Congratulations

[Kun @Kunuid]

@kamellperry_ 基本上没法用。一个简单的问题跑了一个多小时。在上下文长度超过 150K 之后一直 Retry。

[Kun @Kunuid]

@bilibili_create People who follow right now still count as old fans, right? Time for a prize draw!😋

[Kun @Kunuid]

Welcome, Bilibili, I brought you some specialties from your hometown.
One-Click Triple Action.
Do you like it?
@bilibili_create

[bilibili Creator Hub @bilibili_create]

bilibili has always prided itself on community first! so it's very heartening to see that our community has gotten us trending on X. 🥲

Thanks for all the support & let's grow together! ❤️

[Kun @Kunuid]

@doodlestein Is Grok 4.6’s launch‑week 2x usage allowance still active? When will it expire in US time?@grok

[Kun @Kunuid]

@ashfold 真的假的？拉低一下中奖率

[Kun @Kunuid]

@Daniel_Farinax There were indeed numerous "retry" errors today, and I’ve noticed the model’s response quality is poor.

[Kun @Kunuid]

https://t.co/msaLAjXMYL

[Kun @Kunuid]

🚀 Qwen3.8-27B is now open-source! Here’s what developers actually care about 👇

✅ 27B local-friendly size: Runs on a single GPU (16GB+ quantized) — no MoE routing headaches, dead simple to deploy
✅ Inherits 3.8-Max core strengths: Agentic coding, long-horizon autonomous tasks, cowork/office workflows, full multimodal (image + video) upgrades
✅ Controllable thinking: Adjustable reasoning depth + seamless switch between thinking & non-thinking modes
✅ Ultra-long context: Native high context window (extendable to 1M tokens)
✅ Open weights: Direct download on Hugging Face + ModelScope, commercial-friendly

The previous 3.6-27B already punched way above its weight. This 3.8-27B packs flagship-level coding & agent capabilities into a practical, deployable size.

Local deployment, fine-tuning, and agent builders — this one’s for you. Go build!
🔗 Just search Qwen3.8-27B on HF / ModelScope

#Qwen #Qwen3.8 #OpenSourceLLM #LocalAI

[ModelScope @ModelScope2022]

🚀Qwen3.8-27B is officially open-sourced! Apache 2.0 licensed: free to download, deploy, and use commercially. 🤖https://t.co/xr7l4pZDaQ

Your new go-to model for local deployment is here:
🧠 Native multimodal dense architecture. Understands images, text, and video.
🎛️ Flexible thinking control: three levels of reasoning_effort to tune reasoning depth on demand.
📚 262K native context, easily extendable to 1M tokens with YaRN.
🤖 Stronger across coding, office work, and long-horizon agent tasks, with reliable end-to-end delivery.
💻 Runs smoothly on consumer GPUs after quantization. Flagship-level capability, now within everyone's reach.

[Kun @Kunuid]

https://t.co/0sBlI8TuAZ

[Kun @Kunuid]

The official DeepSeek V4-Pro launch announcement is finally here.
Alongside it comes updated API pricing, set to take effect at 16:00 UTC on August 16, 2026 🕒

#DeepSeek #DeepSeekV4Pro #AI #LLM #APIPricing

[DeepSeek @deepseek_ai]

We’re launching DeepSeek-V4-Pro today! 🚀

🔷 Major Agent upgrades with strong production gains!
🔷 Flexible reasoning effort for V4-Pro & V4-Flash: low for simple tasks, high for daily Agent workflows, max for complex tasks.
🔷 Native OpenAI Responses API support, optimized for Codex with one-click setup.

V4 Pro is now available on app/web. Try it via “Expert Mode”.
V4 Pro is also available via API. Model names remain unchanged—please refer to the API docs for setup details.

[Kun @Kunuid]

@notjazii I don't know why, sometimes the High works better than the xhigh .You can try High.