--- title: GLM-5.3: Frontier Coding with Emergent Cyber Capabilities description: Z.ai released GLM-5.3, a post-training-only upgrade over GLM-5.2 that shares the same base model but delivers large gains on coding and long-horizon agentic... canonical: https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: GLM-5.3: Frontier Coding with Emergent Cyber Capabilities | daily.dev og:description: Z.ai released GLM-5.3, a post-training-only upgrade over GLM-5.2 that shares the same base model but delivers large gains on coding and long-horizon agentic... og:url: https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk og:image: https://api.daily.dev/og/posts/WWzXcvxFk.png og:image:alt: GLM-5.3: Frontier Coding with Emergent Cyber Capabilities og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GLM-5.3: Frontier Coding with Emergent Cyber Capabilities **[Hacker News](https://daily.dev/sources/hn)** · 10 min read · 2 upvotes · 1 comments ## Summary Z.ai released GLM-5.3, a post-training-only upgrade over GLM-5.2 that shares the same base model but delivers large gains on coding and long-horizon agentic benchmarks (e.g., Terminal-Bench 3.0 up from 4.6 to 28.3, DeepSWE v1.1 up from 46.2 to 66.9). The release highlights emergent cybersecurity capabilities: on ExploitBench the model's score more than doubled to 54.4%, and in real-world testing with security teams it found 2,436 vulnerabilities across 269 projects, some dating back roughly 40 years, now tracked in a public Z.ai Security Disclosure Ledger. The model runs on the open-source slime RL framework with Megatron and SGLang, achieving over 2.3x higher training throughput. Notably, disabling the 'thinking' mode is no longer supported — applications using thinking.type: disabled must migrate to enabled with reasoning_effort: low before switching to glm-5.3, or requests will fail. Weights will be open-sourced within two weeks. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Do I need to change my API config when upgrading to GLM-5.3 from GLM-5.2? Yes, if the application currently sets thinking.type to disabled, it must be changed to enabled and reasoning_effort set to low before switching the model ID to glm-5.3, otherwise the request will fail. GLM-5.3 no longer supports disabling the thinking mode at all; it only supports low, high, and max reasoning effort levels, with max recommended for coding tasks. _Track breaking API changes like this GLM-5.3 migration requirement on daily.dev before they break production._ ### How much better is GLM-5.3 than GLM-5.2 at finding and exploiting security vulnerabilities? GLM-5.3 shows a large jump in cyber capability: CyberGym rose from 77.2% to 84.5%, ExploitBench more than doubled from 24.4% to 54.4%, and ExploitGym task completions rose from 29 to 105 within a two-hour budget. Testing against real codebases found 2,436 vulnerabilities across 269 projects, with some flaws dating back roughly 40 years, now tracked in the Z.ai Security Disclosure Ledger. _Developers weighing AI-assisted vulnerability discovery can follow benchmarks like these on daily.dev._ ### How does GLM-5.3's coding agent performance compare to Claude Opus 4.8? At High effort, GLM-5.3 reaches 31.4% task completion using around 50K output tokens on Z.ai Code Bench, surpassing Claude Opus 4.8's 29.5% which required 120K tokens. GLM-5.3 still trails Claude Fable 5, which hits 39.5% at Max effort, but delivers stronger token efficiency than both GLM-5.2 and Opus 4.8 at comparable effort levels. _Comparing coding agent efficiency across models is easier when tracking releases like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 126 comments across hackernews (as of 2026-09-14). **TL;DR:** Commenters are impressed that a post-training-only update on the same base model closes much of the gap to closed frontier models at a fraction of the size and cost, while a large side-thread debates whether US labs' cybersecurity guardrails and access restrictions are justified or just gatekeeping/fear-mongering. **Sentiment:** 40% positive · 35% mixed · 25% skeptical **The case for** - Many see it as remarkable that gains come purely from post-training on the same base model, suggesting post-training is now the key lever for progress. - Several note it's impressively close to closed frontier models despite being a much smaller parameter count. - Being free/open and cheap to run is seen as a major practical advantage over closed frontier offerings. - Some praise the release writeup as more honest/researcher-toned than typical marketing hype. - People report already switching from closed models (Opus, GPT) to open models like GLM/Kimi for security-adjacent work because of restrictive guardrails. **The pushback** - Some argue closed frontier models (Fable/Sol/Mythos) still lead, especially at the high end of exploitation-chain benchmarks. - Concerns that quantizing the model locally lobotomizes capability, making local deployment a tradeoff. - Skepticism that post-training gains could just be overfitting to benchmarks rather than real capability. - Criticism that the model isn't yet open-sourced (two-week delay) and that some Chinese labs are moving toward more restricted licenses. - Worry raised about lack of multimodal (vision) support limiting practical use cases. **By community** - hackernews (heated): Excitement about the benchmark gains and openness is real, but the thread is dominated by a heated, sprawling debate over whether closed-lab cybersecurity guardrails are legitimate safety measures or self-serving gatekeeping. **Hottest debate:** Whether the restrictive access to advanced closed-model cybersecurity capabilities reflects genuine government-mandated safety concerns or self-inflicted fear-mongering/gatekeeping by the labs themselves. **Open questions** - Will the promised open-source weight release in two weeks actually happen without further restrictions? - How will consumer hardware (e.g., upcoming Apple silicon) realistically handle running a model this large at usable speed and quality? - Is scaling post-training/RL indefinitely sustainable for continued capability gains, or will it plateau? **Highlights** > In principle I agree but in practice I don’t. The majority of high severity vulnerabilities are not the kind of thing you need a PhD in Comp Sci to comprehend, they are mostly about finding a way to get a system to end up in a state different than was anticipated when entering a particular code path. Exhaustively looking at code and identifying ways to do this is something LLMs are quite good at. They don’t get tired, and you can run them non-stop. They're also (generally) quite good at reading the literal meaning of the code, whereas humans often see the intended meaning first, and can be biased. If you had a tireless junior engineer who was given the job of “make this application get into a state it’s not supposed to be in”, you’d probably get similar results. What Mythos is quite good at is both the first bit and coming up with ways it could chain that together with other bits of unexpected state to create something that forms a meaningful vulnerability rather than a dead end. > — [d1sxeyes on hackernews](https://news.ycombinator.com/item?id=49295824) > the difference is that with open models jailbreaking is trivial if you know what you are doing so this makes a frontier open model infinitely more useful for certain tasks seeing as closed frontier models will just refuse (and jailbreaking them is a waste of time when you have good open models). in some cases (mainly reverse engineering) I have observed GLM 5.2 jailbreaking itself with no effort on my part, the thinking trace revealed that it did some mental gymnastics to pretend it was a crackme or capture the flag competition. > — [teravor on hackernews](https://news.ycombinator.com/item?id=49295155) > Mythos isn't some scary dangerous model that can find high severity bugs seamlessly, that's just Anthropic marketing. Most of the vulnerabilities they found were low severity hyped up to make their model look good, with (I think, maybe?) the exception of a few. Now that Chinese open weight models have similar capabilities, and their guardrails can also just be removed, it doesn't look like anyone has "hacked" into everything because of the scary dangerous models like Anthropic were making it out to be. > — [deepllm on hackernews · 1 comments](https://news.ycombinator.com/item?id=49295339) > I might be just reading my positive bias into that text, but is it possible that it is written less like SV marketing hype trash and more like researchers wrote it? It does feel like it respects both me and my time. Thank you, Z.AI. Amazing what difference it makes when the top of your org are actual university professors. > — [hypfer on hackernews · 1 comments](https://news.ycombinator.com/item?id=49295134) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=49294997) · 300 points · 126 comments ## Community discussion Top comments from developers on daily.dev. **@nerdalytics** · 0 upvotes > This is "just" a post trained model. Therefore, it doesn't support images as in input. Looking strong regardless and will be of good use for sure. --- Tags: [#ai](https://daily.dev/tags/ai), [#ai-agents](https://daily.dev/tags/ai-agents), [#vulnerability](https://daily.dev/tags/vulnerability), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"GLM-5.3: Frontier Coding with Emergent Cyber Capabilities","url":"https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk"},"datePublished":"2026-08-14T05:50:18.124Z","dateModified":"2026-09-14T09:08:59.237Z","description":"Z.ai released GLM-5.3, a post-training-only upgrade over GLM-5.2 that shares the same base model but delivers large gains on coding and long-horizon agentic...","image":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","thumbnailUrl":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"ai,ai-agents,vulnerability,reinforcement-learning","timeRequired":"PT10M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"GLM-5.3: Frontier Coding with Emergent Cyber Capabilities"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk","comment":[{"@type":"Comment","text":"This is “just” a post trained model. Therefore, it doesn’t support images as in input. Looking strong regardless and will be of good use for sure.","datePublished":"2026-08-14T06:22:21.566Z","url":"https://daily.dev/posts/WWzXcvxFk#c-z2L0AE1A3","author":{"@type":"Person","name":"nerdalytics","url":"https://daily.dev/nerdalytics","image":"https://avatars.githubusercontent.com/u/97166791?v=4"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/glm-5-3-frontier-coding-with-emergent-cyber-capabilities-wwzxcvxfk#faq","mainEntity":[{"@type":"Question","name":"Do I need to change my API config when upgrading to GLM-5.3 from GLM-5.2?","acceptedAnswer":{"@type":"Answer","text":"Yes, if the application currently sets thinking.type to disabled, it must be changed to enabled and reasoning_effort set to low before switching the model ID to glm-5.3, otherwise the request will fail. GLM-5.3 no longer supports disabling the thinking mode at all; it only supports low, high, and max reasoning effort levels, with max recommended for coding tasks. Track breaking API changes like this GLM-5.3 migration requirement on daily.dev before they break production."}},{"@type":"Question","name":"How much better is GLM-5.3 than GLM-5.2 at finding and exploiting security vulnerabilities?","acceptedAnswer":{"@type":"Answer","text":"GLM-5.3 shows a large jump in cyber capability: CyberGym rose from 77.2% to 84.5%, ExploitBench more than doubled from 24.4% to 54.4%, and ExploitGym task completions rose from 29 to 105 within a two-hour budget. Testing against real codebases found 2,436 vulnerabilities across 269 projects, with some flaws dating back roughly 40 years, now tracked in the Z.ai Security Disclosure Ledger. Developers weighing AI-assisted vulnerability discovery can follow benchmarks like these on daily.dev."}},{"@type":"Question","name":"How does GLM-5.3's coding agent performance compare to Claude Opus 4.8?","acceptedAnswer":{"@type":"Answer","text":"At High effort, GLM-5.3 reaches 31.4% task completion using around 50K output tokens on Z.ai Code Bench, surpassing Claude Opus 4.8's 29.5% which required 120K tokens. GLM-5.3 still trails Claude Fable 5, which hits 39.5% at Max effort, but delivers stronger token efficiency than both GLM-5.2 and Opus 4.8 at comparable effort levels. Comparing coding agent efficiency across models is easier when tracking releases like this on daily.dev."}}]} ```