Releases·Model Release
GLM-5.2: Open Weights, Near-Frontier Intelligence — Zixuan Li, Z.ai
Z.ai head Zixuan Li introduced GLM-5.2 in a talk, positioning the open-weight model between Claude Opus 4.7 and 4.8 on long-horizon coding and agentic benchmarks. He argued open weights let enterprises, governments, and fine-tuning customers run and shape models on their own infrastructure. He also announced Z Code, Z.ai's own coding harness built for GLM-5.2 and compatible with other frontier models.
28 engagement1 sourceSun, Sep 27, 2026, 05:00 PM
AI Overview
- Z.ai head Zixuan Li says GLM-5.2 scores between Claude Opus 4.7 and 4.8 on hard long-horizon coding and agentic benchmarks.
- GLM-5.2's non-thinking mode beats GLM-5.1 with thinking on, and it leads open-weight models on the Artificial Analysis Intelligence Index.
- Li says open weights let enterprises and governments self-host, fine-tune for law, finance and security, and co-design the roadmap.
- The GLM name comes from a 2021 paper placing Z.ai among the earliest LLM labs.
- Z.ai also launched Z Code, a coding harness built for GLM-5.2 that works with other frontier models.
Related
Developments·Paper
SaferAI report: Z.ai's GLM-5.2 narrows frontier gap but lags on safety
A new SaferAI report finds Z.ai's open-weight GLM-5.2 is only months behind frontier models on cyber and bio capabilities, but its safety practices lag significantly. The evaluation, run via Z.ai's public API, shows GLM-5.2 refused none of the dangerous requests tested, highlighting a growing divide between capability and safety.
A new SaferAI report finds Z.ai's open-weight GLM-5.2 is only months behind frontier models on cyber and bio capabilities, but its safety practices lag significantly. The evaluation, run via Z.ai's public API, shows GLM-5.2 refused none of the dangerous requests tested, highlighting a growing divide between capability and safety.
0 engagement1 sourcerssTue, Aug 4, 2026, 08:05 PM
Releases·Model Releasefeatured
Z.ai releases GLM-5.3 with cyber and coding gains, holds open weights for safety review
Z.ai announced GLM-5.3, an open-weight model built on the same mixture-of-experts base as GLM-5.2, with all gains from extended post-training. It scores 84.5% on CyberGym, edging Claude Mythos 5 and GPT-5.6 Sol, and shows large jumps on Terminal-Bench 3.0 and DeepSWE v1.1. The company is holding back weights for a cyber safety review.
Z.ai announced GLM-5.3, an open-weight model built on the same mixture-of-experts base as GLM-5.2, with all gains from extended post-training. It scores 84.5% on CyberGym, edging Claude Mythos 5 and GPT-5.6 Sol, and shows large jumps on Terminal-Bench 3.0 and DeepSWE v1.1. The company is holding back weights for a cyber safety review.
Updated Tue, Aug 18, 2026, 08:46 PM — API went live on 2026-08-18, priced same as GLM-5.2, via official API and partner gateways.
1416.1k engagement6 sourcesrss, reddit, xFri, Aug 14, 2026, 12:00 AM
Releases·Model Releasefeatured
Z.ai releases GLM-5.3-Flash, first open-weight glm5_next model
Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series and the first open-weight release of the glm5_next architecture. The model is positioned as outperforming GLM-5.2 at one-tenth the price while approaching Claude Opus 4.8 on coding and agentic benchmarks. It introduces a hybrid sparse and linear architecture.
Z.ai released GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series and the first open-weight release of the glm5_next architecture. The model is positioned as outperforming GLM-5.2 at one-tenth the price while approaching Claude Opus 4.8 on coding and agentic benchmarks. It introduces a hybrid sparse and linear architecture.
42028.4k engagement6 sourcesreddit, x, githubWed, Aug 26, 2026, 03:17 PM
Releases·Model Release
Writer launches Palmyra X6, a GLM-5.2-based model to cut token costs
Writer introduced Palmyra X6, a new flagship model built as a post-training variation of Z.ai's open-source GLM-5.2, aiming to provide deployment-ready capabilities at lower per-token costs. The launch responds to growing industry pressure to reduce AI deployment expenses.
Writer introduced Palmyra X6, a new flagship model built as a post-training variation of Z.ai's open-source GLM-5.2, aiming to provide deployment-ready capabilities at lower per-token costs. The launch responds to growing industry pressure to reduce AI deployment expenses.
0 engagement2 sourcesrss, arxivThu, Aug 13, 2026, 09:13 PM
News·Company News
Mistral hosts rival Z.ai's GLM-5.2, priced below its own flagship
Mistral began hosting Z.ai's GLM-5.2 on its platform, pricing it cheaper than its own flagship Mistral Medium 3.5. The move is notable because the two companies are competitors, and it may signal a strategic shift toward selling compute and specialized models rather than pursuing frontier development.
Mistral began hosting Z.ai's GLM-5.2 on its platform, pricing it cheaper than its own flagship Mistral Medium 3.5. The move is notable because the two companies are competitors, and it may signal a strategic shift toward selling compute and specialized models rather than pursuing frontier development.
152 engagement1 sourceredditFri, Aug 14, 2026, 02:36 AM