Z.ai's GLM-5.2 Reaches Together AI and OpenRouter, Tops DeepSWE Open-Source Leaderboard at 44%
Chinese AI lab Z.ai's GLM-5.2, an AI model released with public model weights, is moving from benchmark leaderboards into wider deployment on hosted services and local setups. Together AI is serving the model on OpenRouter, a free US-hosted chat app is running on Together's platform, and a 467GB build has been uploaded for local use. DeepSWE, a software-engineering benchmark, put GLM-5.2 at the top of its open-source leaderboard with a 44% pass rate at max effort, 17 percentage points ahead of Kimi K2.7 Code.
The rollout extends GLM-5.2's reach beyond the benchmark streak that had already made it the highest-ranked open-weight model on Artificial Analysis's Intelligence Index and the only one above 60% on Vals AI's Vibe Code Bench, a test of whether models can build web applications from scratch. Code Arena's frontend leaderboard later placed GLM-5.2 Max second, 29 points ahead of Claude Opus 4.7 (Thinking) and behind only Anthropic's Claude Fable 5.
From the sources (25 posts)
@artificialanlys@jeremyphoward @Zai_org GLM-5.2 is between GPT-5.5 and Opus 4.8 in our new agentic knowledge work eval that we released just today. Very impressive
@nandodfRT @rasbt: Just caught up with the recent GLM-5.2 release. The best open-weight model today. Architecture-wise, it's build on the GLM-5 an…
@_arohan_RT @jeremyphoward: Wow. @Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and…
@andrewcurran_Rave reviews all over my timeline for GLM 5.2 the last 48 hours, open source is having an exceptional month.
@clementdelangue“Cost per task varies by ~800x across models tested: Claude Fable 5 leads the benchmark but costs more than $31 per task on average, compared to ~$0.04 for DeepSeek V4 Flash (max). The strongest price/performance options are open weights mo
@chris_j_paxtonRT @jietang: GLM-5.2 is Fully Open, Frontier Intelligence Belongs to Everyone Today, the sudden restriction of certain frontier models is…
@clementdelangueRT @jeremyphoward: Wow. @Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and…
@ollamaRT @jeremyphoward: Wow. @Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and…
@clementdelangueRT @AndrewCurran_: Rave reviews all over my timeline for GLM 5.2 the last 48 hours, open source is having an exceptional month.
@zai_orgRT @ArtificialAnlys: Announcing AA-Briefcase, the benchmark for the next era of agentic knowledge work AA-Briefcase is our new benchmark f…
@andersonbcdefgRT @teortaxesTex: ok, here's how GLM 5.2 performs on a bench it definitely didn't see, and where GLM 5.1 scored 0.0%. Closer to Opus 4.8 th…
@matvellosoAll day using GLM 5.2. Didn't miss much. First open model that passes the bar as a daily driver. Things are not going to be the same. Damn, now I want to buy some serious hardware.
@wesrothRT @WesRoth: GLM-5.2 Max reached second place in the Code Arena Frontend leaderboard, behind only the currently unavailable Claude Fable 5.…
@quixiaiRT @jeremyphoward: Wow. @Zai_org GLM 5.2 is a marvel! It is *at least* as good as Opus 4.8 and GPT 5.5. It's super fast, inexpensive, and…
@zai_orgRT @ZixuanLi_: GLM-5.2 delivers a substantial leap in app development capabilities, which also represent demanding long-horizon tasks. Res…
@zai_orgLong-horizon is more than a concept. It should live in real-world scenarios, empowering AI builders to solve the problems that matter. And more scenarios are on the way.
@teortaxestexZhipu post-training team is very very good
@zai_orgRT @CunxiangWang: GLM-5.2 is not only stronger on benchmarks, but also much better in real app development scenarios — iOS, Android, WeChat…
@teortaxestexGLM results are insanely high If this is "distillation", it's doing better than any previous attempt Eg ProofBench should be amenable both to distillation and RLVR, yet V4, Kimi, Mimo, Grok lol are all hopeless. Zhipu will obviously aim to
@teortaxestexRT @banteg: not a bad result from glm-5.2, it found this after gpt-5.5 xhigh
@quixiaiRT @AlexFinn: I can't believe this is real I have GLM 5.2 running 100% locally on my Mac Studio. 2 bit quant. The results I'm getting are…
@vipulvedSpent the last 3 hours playing with GLM-5.2. It’s truly a stunning model. What a time to be alive.
@ollamaRT @matvelloso: All day using GLM 5.2. Didn't miss much. First open model that passes the bar as a daily driver. Things are not going to be…
@matansfRT @alphatozeta8148: @baseten model performance team is absolutely cracked. @Zai_org GLM 5.2 is now 4x faster running at full 1M context!…
@teortaxestexInteresting that in "GLM 5.2 is on par with X model", X is so widely distributed over different capabilities, and it's not like the gap is just greater for the difficult ones. It plainly can't do some parlor tricks Sonnet 3.5 pulled off. In