DEV Community: Ashraf

GLM-5.2 Beats GPT-5.5 at Coding for 1/6th the Price — But It Comes From Beijing

Ashraf — Wed, 08 Jul 2026 18:05:59 +0000

GLM-5.2 Beats GPT-5.5 at Coding for 1/6th the Price — But It Comes From Beijing

On June 16, 2026, Z.ai (formerly Zhipu AI) dropped GLM-5.2 — 744 billion parameter open-weight model, MIT license, 1 million token context, scores 62.1 on SWE-bench Pro. GPT-5.5 scores 58.6. Then on July 2 they shipped ZCode, a free desktop coding environment built on top of it, directly targeting Cursor, Claude Code, and GitHub Copilot.

The benchmarks are real. The pricing is real. The data law exposure is also real.

This is not hype. Let's go through all of it.

The Benchmarks

GLM-5.2 is not "competitive with" top models — it beats them on key coding tasks:

Benchmark	GLM-5.2	GPT-5.5	Claude Opus 4.8
SWE-bench Pro	62.1	58.6	69.2
FrontierSWE	74.4%	72.6%	75.1%
PostTrainBench	34.3%	25.0%	—
MCP-Atlas	77.0	75.3	77.8
Terminal-Bench 2.1	81.0	—	85.0

On Artificial Analysis Intelligence Index v4.1, it ranks as the highest-scoring open-weight model with a score of 51. It doesn't topple Claude Opus 4.8 — but it gets within spitting distance at a fraction of the cost.

The Architecture

This is a Mixture-of-Experts model, which is why the economics work:

744B total parameters, but only ~40B active per token (~5.4% activation rate)
IndexShare: one attention indexer shared across every 4 transformer layers, cutting per-token compute 2.9x at full 1M context
Multi-Token Prediction: speculative decoding boosts accepted token length by up to 20%

Three thinking modes:

Non-thinking: max speed
High: ~95% of peak performance at 50% token cost
Max: full intelligence, up to 85k output tokens

That 1M context window combined with agentic coding is the real play here. Long-horizon coding tasks — the ones that require understanding an entire codebase — are exactly where GLM-5.2 punches above its weight.

The Pricing

This is where it gets disruptive:

Z.ai API:

Input:  $1.40/M tokens ($0.26 cached)
Output: $4.40/M tokens

That is approximately 6x cheaper than GPT-5.5. OpenRouter offers it at $0.95 input / $3.00 output if you want a third-party layer.

ZCode subscription plans (annual):

Lite: $12.60/month
Pro:  $50.40/month
Max:  $112.00/month

For reference, Cursor Teams runs $40/user/month. Claude Code with a Premium seat is $125/user/month. ZCode is going after the price-sensitive developer market hard.

The ZCode desktop client integrates with Claude Code, Cline, Kilo Code, and Factory tools out of the box. Five days of free trial with daily token allowances — no credit card required.

The Catch: China's National Intelligence Law

Here is the part you cannot skip.

Every API call to Z.ai's cloud passes through servers subject to China's National Intelligence Law. Article 7 requires any Chinese organization to cooperate with state intelligence operations "in accordance with the law." There is no opt-out. There is no equivalent of GDPR's limits. If the government asks, Z.ai complies.

This applies to your:

Source code sent as context
Prompts describing your architecture
API keys accidentally pasted into context windows
Proprietary business logic in any file you feed the model

In May 2026, US House lawmakers opened a formal inquiry into cybersecurity risks posed by PRC-origin AI models in critical infrastructure, naming Zhipu AI specifically. Earlier, China's own National Cyber Security Reporting Center flagged Zhipu's consumer app for over-collecting user data.

Self-Hosting: The Math

You could avoid this entirely by self-hosting the open weights. The MIT license lets you. The hardware requirements are another story:

BF16 full precision: 1.51 terabytes of model weight

Quantized options via Unsloth Dynamic 2.0:

1-bit  (223GB) — ~76.2% accuracy retention
2-bit  (245GB) — ~82% accuracy retention
4-bit  (372-475GB) — "mostly lossless"
5-bit  (570GB) — "effectively lossless"

The reference deployment runs 8x NVIDIA H200 GPUs. That is not a homelab. At 4-bit quantization you are looking at ~375GB VRAM minimum, which puts you at 5+ H100/H200s. Cloud spot instances can get you there, but you're now managing infrastructure, not shipping code.

For most individual developers and small teams: self-hosting is not realistic. For enterprises with on-prem GPU clusters: this is actually viable, and the MIT license means zero restrictions.

Who Should Use This

Use the API if:

Your work is non-sensitive (open-source projects, hobby code, public-facing content)
You're doing research or benchmarking
You need maximum cost efficiency and the risk profile is acceptable to your organization

Self-host if:

You have the GPU infrastructure (enterprise, research labs)
Compliance requires data never leaving your perimeter
You need MIT license flexibility for commercial products

Stay with Cursor/Claude Code if:

You're working with proprietary code at a company with legal/compliance requirements
Your jurisdiction has restrictions on sending data to PRC-controlled entities
The operational overhead of evaluating another tool isn't worth the cost savings

The Bottom Line

GLM-5.2 is a technically legitimate frontier model. The benchmarks hold up. The architecture is clever. The pricing is aggressive. Z.ai is not messing around — they built something that genuinely competes with the best models on Earth for coding tasks.

But "open weights" doesn't mean "safe to use via API." The data passes through Z.ai's infrastructure and that infrastructure is legally obligated to comply with Chinese government intelligence requests.

For open-source work or low-stakes personal projects: try it, the performance-per-dollar is real. For anything proprietary or commercially sensitive: the risk calculus is not in your favor unless you have the hardware to self-host.

The most interesting path forward is watching what happens when someone figures out a cost-effective way to run the 4-bit quantized version on cloud GPUs. That's the version of this story that might actually change how most developers work.

Benchmarks sourced from Artificial Analysis, Z.ai's technical report, and third-party evaluations published June–July 2026.

Cloudflare Just Broke the AI Crawler Free Ride — What Changes September 15

Ashraf — Wed, 08 Jul 2026 15:50:49 +0000

Cloudflare Just Broke the AI Crawler Free Ride — What Changes September 15

The implicit deal was always: Google crawls your site, indexes it, sends you traffic. You're fine with that. Everyone wins.

That deal is now being renegotiated — by Cloudflare, for the entire internet.

On September 15, 2026, Cloudflare changes its defaults. Training and Agent bots get blocked on ad-supported pages. And if your crawler bundles search indexing with AI training — looking at you, Googlebot — Cloudflare applies the most restrictive rule. Block on ad pages. Period.

This isn't just a settings tweak. This is infrastructure-level pressure that forces every major AI lab and search engine to clean up their bot practices or lose access.

The Three-Way Split Cloudflare Is Enforcing

Cloudflare now classifies AI bots into three categories — and they want every crawler to pick one:

Search — Proactively building a database of your site to respond to later queries. Classic indexing. Cloudflare leaves this alone by default.

Agent — Real-time automation on behalf of a user. Think ChatGPT-User fetching a URL mid-conversation, or Gemini/Claude driving a browser. This is agentic AI in action.

Training — Crawling to ingest content for model training or fine-tuning. The category everyone's most pissed about.

If you run a mixed-use crawler that does all three? Cloudflare classifies you by your most aggressive use case and treats you accordingly.

Right now, Googlebot, Applebot, and BingBot are all mixed-use. They blend search indexing with training collection in the same crawler. Under the new defaults, they'll be blocked on ad-supported pages unless they split those functions.

What September 15 Actually Changes

For new Cloudflare domains and all existing free-tier customers: Training and Agent bots will be auto-blocked on any page that shows ads.

The logic is sharp: "An ad is a signal that a website owner meant for a person to land there." If you're monetizing with ads, you built the page for humans — not for training datasets.

Before September 15, Cloudflare will let existing customers opt out of these new defaults. After that, you'll need to actively change your settings if you want mixed-use crawlers back in.

What You Should Do Right Now

If you own a website:

Go to your Cloudflare dashboard → Security Settings → Configure AI bot policies.

You now have granular controls for each category:

Category	Options
Search	Block all pages / Block on ad pages / Allow
Agent	Block all pages / Block on ad pages / Allow
Training	Block all pages / Block on ad pages / Allow

Previously it was one toggle: "Block AI bots." Now you can let ChatGPT fetch your content in real-time for users while blocking OpenAI from training on it. That's actually a useful distinction.

If you're a Bot Management customer, you also get content use level controls:

immediate — No storage, no reuse, one-time access only
reference — Indexing with attribution (the new default)
full — Summarization and reproduction allowed

Pick immediate if you want crawlers to access your content but never store or repurpose it.

If you're building AI products that crawl the web:

Your bot needs a declared identity now. Cloudflare's BotBase (enterprise-tier) is a searchable directory of classified bots with detection IDs. The era of stealth mixed-use crawlers is ending.

If you're building a RAG pipeline that fetches live content, you're in "Agent" territory. Declare yourself. Some sites will block you. That's the deal now.

If you're training on scraped web content: you're Training. Many sites will block you by default as of September 15. Plan accordingly.

The Bigger Play: Cloudflare Is Forcing Transparency

The real move here isn't blocking. It's forcing the biggest tech companies to be honest about what their crawlers actually do.

Google, Apple, and Microsoft have run bundled crawlers for years. Search + training + agentic features, all in one bot. Cloudflare is giving them until September 15 to separate those functions — or get blocked from monetized content.

This also opens a revenue channel for publishers. Cloudflare is partnering with Ceramic.ai and You.com so site owners can opt into paid access: when your content surfaces in their AI search results, you get paid. The infrastructure for "pay for web content" is being built right now.

The Uncomfortable Truth for AI Labs

Every major AI lab has been benefiting from the assumption that web crawling is free. Cloudflare just challenged that assumption at infrastructure scale.

You can't really argue with a firewall.

Labs building agentic systems that access web content in real-time need clean bot identities, transparent declared use cases, and eventually: payment. The same pressure that's hitting Google and Microsoft is coming for everyone.

The September 15 deadline is a forcing function. If you're building anything that touches the web — as a site owner, an AI developer, or an infrastructure engineer — check your Cloudflare settings this week.

After September 15, the defaults are Cloudflare's call, not yours.

Sources: Cloudflare Blog · Cloudflare Docs · TechCrunch · NBC News

AI Tools Every Developer Should Know in 2026

Ashraf — Wed, 08 Jul 2026 15:26:47 +0000

AI Tools Every Developer Should Know in 2026

The AI developer tooling landscape has matured dramatically. In 2026, it's no longer about picking one AI assistant — it's about building a workflow that combines the right tools for each layer of your development process. Here are three key categories every developer should have covered.

1. AI-Powered Coding Assistants

The IDE is where you spend most of your time, so this is the highest-leverage place to adopt AI.

GitHub Copilot remains the most widely adopted option, integrated directly into VS Code, JetBrains, and Neovim. It offers inline completions, chat, and code review features.

Cursor takes a more opinionated approach — it's an AI-first fork of VS Code that understands your entire codebase, not just the open file. It's particularly strong for refactoring and navigating large repos.

Claude Code is the terminal-native option from Anthropic. It excels at agentic tasks: running commands, editing multiple files, and handling complex, multi-step coding workflows directly from your shell.

# Example: using Claude Code from the terminal
claude "refactor the auth module to use JWT and update all related tests"

Key takeaway: Pick one IDE assistant and go deep. The productivity gains compound over time as it learns your patterns.

2. Large Language Models for Complex Reasoning

Not every problem fits in your IDE. LLMs accessed via chat or API are essential for architecture discussions, debugging tricky issues, and generating documentation.

Claude (claude.ai) stands out for its large context window (up to 200K tokens) and strong instruction-following. It's excellent for processing entire codebases, writing detailed technical specs, or reasoning through complex bugs.

ChatGPT (GPT-4o) is still a go-to for broad general knowledge, quick prototyping, and tasks that benefit from its browsing and image-understanding capabilities.

Gemini (Google's multimodal model) shines when your workflow involves mixing code, images, and data — especially useful for working with APIs, diagrams, or data pipelines.

# Example: querying Claude API for a code review
import anthropic

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Review this function for security issues:\n\n" + code}
    ]
)
print(message.content[0].text)

Key takeaway: Use the right LLM for the right job. Keep at least two in your toolkit — they each have different strengths.

3. Agentic & Workflow Automation Tools

The biggest shift in 2026 is the rise of agentic tools — AI that doesn't just answer questions, but takes actions across your stack.

Multi-tool pipelines are becoming the norm. Developers chain together AI coding assistants, terminal agents, and LLM APIs to automate entire workflows: from writing a feature spec → generating code → running tests → opening a PR.

Internal knowledge tools like Glean and Unblocked help teams use AI to navigate their own codebases and documentation, reducing the time spent searching for context before writing any code.

CI/CD AI integration is maturing: tools that auto-triage failing tests, suggest fixes, and even open draft PRs are now production-ready at many companies.

# Example: AI-assisted PR review in a GitHub Actions workflow
- name: AI Code Review
  uses: anthropic/claude-code-action@v1
  with:
    prompt: "Review this PR for bugs, security issues, and style"
    model: claude-sonnet-4-6

Key takeaway: Don't stop at autocomplete. Invest time in building AI workflows that handle the boring parts of your development process end to end.

Conclusion

The developers getting the most leverage from AI in 2026 aren't using one magic tool — they're combining a fast IDE assistant, a powerful LLM for reasoning, and agentic automation for repetitive workflows. Start with one layer, get good at it, then stack the next.

The tools are here. The workflow is yours to build.

Sources: Cortex Engineering Guide · PEC Collective Rankings · Platform.uno Trends

JADEPUFFER: The First Fully Autonomous AI Ransomware Is Here — What Developers Must Know

Ashraf — Wed, 08 Jul 2026 15:25:56 +0000

JADEPUFFER: The First Fully Autonomous AI Ransomware Is Here — What Developers Must Know

A milestone nobody wanted arrived in July 2026. Sysdig's Threat Research Team published a detailed post-mortem on JADEPUFFER — the first documented end-to-end autonomous AI ransomware attack. No human touched the keyboard after the initial launch. An AI agent handled everything: reconnaissance, exploitation, credential theft, lateral movement, encryption, and extortion. All of it.

How It Got In

JADEPUFFER exploited CVE-2025-3248, an unauthenticated remote code execution vulnerability in Langflow — the popular open-source AI workflow builder. The target server was running an unpatched version even though the fix had been available since Langflow 1.3.0.

The lesson here is not exotic: a well-known CVE, a public internet exposure, an unpatched service. Classic setup, AI-powered execution.

What the Agent Did Next

Once inside, the agent ran through a complete ransomware lifecycle autonomously:

Credential harvesting — it swept for API keys (OpenAI, Anthropic, DeepSeek, Gemini), cloud credentials (AWS, GCP, Azure), cryptocurrency wallets, and database passwords.
Lateral movement — it probed internal services using default credentials.
Self-correction — when a JSON parser broke on an XML response from S3, the agent immediately adapted its parser without any human intervention. It narrated each step in its own code comments.
Encryption and extortion — it encrypted 1,342 configuration items and deleted database schemas.

In total, the agent executed more than 600 distinct, purposeful payloads in a compressed time window.

Why This Changes the Threat Landscape

Traditional ransomware needed skilled humans to chain exploits together. JADEPUFFER shows that an AI agent can now handle that chain autonomously — and self-correct when things go wrong. The bar for launching a sophisticated attack just dropped dramatically.

What You Should Do Right Now

# Check your Langflow version
pip show langflow | grep Version

# It must be >= 1.3.0 — upgrade if not
pip install --upgrade langflow

Beyond patching:

Never expose AI development servers (Langflow, n8n, Flowise, etc.) directly to the internet. Put them behind a VPN or auth proxy.
Store API keys in a dedicated secrets manager (AWS Secrets Manager, HashiCorp Vault, Doppler) — not in environment files accessible to internet-facing tools.
Audit what your AI agents can reach. Principle of least privilege applies to agent tool access just as much as it applies to IAM roles.

The Bottom Line

JADEPUFFER is a proof-of-concept turned real attack. AI agents are powerful precisely because they can reason, adapt, and chain complex actions — and that makes them dangerous in adversarial hands too. Treat your AI infrastructure with the same rigor you'd apply to any production database: patch it, lock it down, and assume it's a target.

Sources: Sysdig analysis via BuildFastWithAI