Security research often involves questions that general-purpose models refuse Meet Dolly 1 Security and Dolly 2 Security, uncensored variants of GLM-5.3 Flash and GLM-5.3 for defensive testing and red-team workflows. Run them through the Zro API or CLI with zero request retention. Since our July launch, we’ve also added new models, agent integrations, yearly plans, and referrals.
Hi Product Hunt! We launched Zro here in July for fast, private inference for coding agents. Since then, we’ve served over 250B tokens. Today we’re expanding into private models for cybersecurity research.
General-purpose models can refuse those questions before the investigation gets anywhere. Dolly Security is our answer.
Live on Zro today:
Dolly 1 Security: Based on GLM-5.3 Flash. $0.45 input / $1.50 output per 1M tokens.
Dolly 2 Security: Based on GLM-5.3, abliterated in-house with a focus on retaining general capability. $2.10 input / $6.30 output per 1M tokens. We’ve also published its weights and evaluation.
Product Hunt offer: 25% off your first month of Pro or Max with ZROPRODUCTHUNT.
Referrals: Give a friend $5 in usage credit and get $5 when they start a paid subscription.
Community:Join our Discord to talk with the team and other builders.
What we’re working on next: Kimi K3.1 on our stack and a security-focused version of DeepSeek V4.1 Flash. We’re also working toward stronger evaluations and clearer evidence of how our security models are modified.
We’ll be here today to answer questions and hear what you think.
The published weights and evaluation for Dolly 2 is the part that stood out to me — abliterating while trying to retain general capability is easy to claim and hard to show. I'm curious how you check that in practice: do the evals you published cover both the security tasks and ordinary coding/general tasks side by side, so a regression in one shows up before a release?
The evaluation is the part I'd spend time on here. I'd want to see security-task accuracy and ordinary coding accuracy side by side after the model changes.
Hey Product Hunt, I'm Emir from the Zro team 👋
We launched here in July with one goal: fast, private inference for coding agents.
We're a small team that likes digging into how inference actually works. We've sent a bunch of PRs upstream to SGLang and vLLM, and we write up what we find along the way. See this
Today's launch adds security models and a free plan.
Would love to hear what you're building with coding agents and what you'd want from us next. I'm around all day 🙂
About Zro: Uncensored Security Models on Product Hunt
“Abliterated open models for security work”
Zro: Uncensored Security Models launched on Product Hunt on October 28th, 2026 and earned 115 upvotes and 13 comments, placing #10 on the daily leaderboard. Security research often involves questions that general-purpose models refuse Meet Dolly 1 Security and Dolly 2 Security, uncensored variants of GLM-5.3 Flash and GLM-5.3 for defensive testing and red-team workflows. Run them through the Zro API or CLI with zero request retention. Since our July launch, we’ve also added new models, agent integrations, yearly plans, and referrals.
Zro: Uncensored Security Models was featured in API (98.8k followers), Developer Tools (522.1k followers) and Artificial Intelligence (482.4k followers) on Product Hunt. Together, these topics include over 236.8k products, making this a competitive space to launch in.
Who hunted Zro: Uncensored Security Models?
Zro: Uncensored Security Models was hunted by Ben Lang. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Zro: Uncensored Security Models stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Ben Lang
Hi Product Hunt! We launched Zro here in July for fast, private inference for coding agents. Since then, we’ve served over 250B tokens. Today we’re expanding into private models for cybersecurity research.
General-purpose models can refuse those questions before the investigation gets anywhere. Dolly Security is our answer.
Live on Zro today:
Dolly 1 Security: Based on GLM-5.3 Flash. $0.45 input / $1.50 output per 1M tokens.
Dolly 2 Security: Based on GLM-5.3, abliterated in-house with a focus on retaining general capability. $2.10 input / $6.30 output per 1M tokens. We’ve also published its weights and evaluation.
Free plan: Daily usage on Flash models.
Try Dolly 2 with Claude Code
More from this launch
Product Hunt offer: 25% off your first month of Pro or Max with ZROPRODUCTHUNT.
Referrals: Give a friend $5 in usage credit and get $5 when they start a paid subscription.
Community: Join our Discord to talk with the team and other builders.
What we’re working on next: Kimi K3.1 on our stack and a security-focused version of DeepSeek V4.1 Flash. We’re also working toward stronger evaluations and clearer evidence of how our security models are modified.
We’ll be here today to answer questions and hear what you think.