GLM-5.2

GLM-5.2 is an open-weight large language model developed by the Chinese company Zhipu AI, which sells its products internationally under the brand Z.ai. Released in June 2026 under the permissive MIT License, it is a Mixture-of-Experts (MoE) model with roughly 753 billion total parameters and about 40 billion active per token, a 1 million token context window, and a focus on agentic coding and long-horizon task execution. At launch it became the highest-scoring open-weight model on the Artificial Analysis Intelligence Index, and it out-scored OpenAI's GPT-5.5 on several long-horizon coding benchmarks at a fraction of the price. ^[1][2][3]

Scroll sideways for more →

What is GLM-5.2?

GLM-5.2 is the third release in Zhipu AI's fifth-generation GLM (General Language Model) line, following GLM-5 and GLM-5.1. It is built for "long-horizon" work: multi-step reasoning, tool use, and software-engineering tasks that unfold over many turns and large codebases. Z.ai describes it as a substantial leap in long-horizon capability over GLM-5.1, delivered for the first time on a full 1 million token context. ^[2] The weights are published on Hugging Face under the repository zai-org/GLM-5.2, so anyone can download, run, fine-tune, and deploy the model commercially. ^[3]

Who developed GLM-5.2?

GLM-5.2 was built by Zhipu AI, a Beijing-based AI company founded in 2019 as a spin-off from Tsinghua University's Knowledge Engineering Group. The company markets its models and API internationally under the Z.ai brand and is publicly listed in Hong Kong (HKEX: 2513). GLM-5.2 continues Zhipu's strategy of shipping frontier-class models with open weights, positioning the firm as a direct competitor to both Western labs and other Chinese open-weight developers. ^[1][8]

How does GLM-5.2 relate to GLM-5 and GLM-5.1?

The series began with GLM-5, a 744 billion parameter open-weight flagship released on February 11, 2026. GLM-5.1 followed on April 7, 2026 as an incremental update. GLM-5.2 is the coding and agent focused refinement of that lineage. It keeps the MoE design but raises the context window from 200,000 tokens to 1 million and adds architectural changes aimed at cheaper long-context inference. Z.ai reports large coding gains over GLM-5.1, for example Terminal-Bench 2.1 rising to 81.0 from 63.5. ^[2][1]

What is GLM-5.2's architecture?

According to its Hugging Face config.json, GLM-5.2 uses the GlmMoeDsaForCausalLM architecture (model type glm_moe_dsa). It has 78 transformer layers, a hidden size of 6,144, and 64 attention heads. The MoE feed-forward network contains 256 routed experts plus 1 shared expert, with 8 routed experts activated per token, giving roughly 40 billion active parameters out of about 753 billion total. Artificial Analysis lists the model slightly lower, at 744 billion total and 40 billion active. The vocabulary is 154,880 tokens and the weights ship in bfloat16. ^[4][2][1]

Two efficiency features distinguish the model. "IndexShare" reuses a single lightweight sparse-attention indexer across every four layers, which Z.ai says cuts per-token compute by about 2.9 times at the full 1 million token context. An upgraded Multi-Token Prediction layer improves speculative decoding, raising accepted token length by up to 20 percent. Together with the deep sparse attention (DSA) design implied by the architecture name, these changes target the cost of very long context inference. ^[2]

Is GLM-5.2 open source?

Yes, in the open-weight sense. GLM-5.2's weights are released under the MIT License, one of the most permissive licenses available, allowing unrestricted commercial use, modification, and redistribution with no regional limits. ^[1][3] You can self-host by downloading the weights from Hugging Face, and community and vendor quantizations (including NVIDIA NVFP4, GGUF, and Apple MLX MXFP4 builds) make it runnable on a range of hardware. It is also served through Z.ai's own API and third-party inference providers. As with most open-weight models, the weights and license are public, but the full training data and pipeline are not. ^[3]

How does GLM-5.2 perform on benchmarks?

GLM-5.2 posts strong scores across reasoning, math, and coding. On Z.ai's reported evaluations it reaches 99.2 on AIME 2026, 92.5 on HMMT February 2026, 91.2 on GPQA-Diamond, and 40.5 on Humanity's Last Exam (54.7 with tools). ^[2] Independent testing by Artificial Analysis placed GLM-5.2 at 51 on its Intelligence Index (v4.1), the top score for any open-weight model at release, ahead of DeepSeek V4 Pro and MiniMax-M3 (both 44) and Kimi K2.6 (43). Artificial Analysis also noted that the model is unusually verbose, using around 43,000 output tokens per task, which raises real-world cost despite the low per-token price. ^[1]

How good is GLM-5.2 at coding?

Coding and agentic tool use are GLM-5.2's headline strengths. It scores 62.1 on SWE-bench Pro and 74.4 on FrontierSWE, both of which, according to Z.ai and reporting by VentureBeat, beat GPT-5.5 (58.6 and 72.6 respectively). On Terminal-Bench 2.1 it reaches 81.0 (82.7 best reported), roughly 18 points above GLM-5.1. ^[2][6] Zhipu also shipped a dedicated agent harness alongside the model, explicitly targeting Anthropic's Claude Code style workflows, and the model has been integrated into more than 20 coding tools. ^[8]

How does GLM-5.2 compare to DeepSeek V4 and Kimi K2.6?

Among open-weight peers released in 2026, GLM-5.2 leads the Artificial Analysis Intelligence Index at 51, versus 44 for DeepSeek V4 Pro and 43 for Kimi K2.6 from Moonshot AI. ^[1] DeepSeek V4 is an open-weight MoE family aimed at broad general capability, while Kimi K2.6 is a trillion-parameter MoE tuned for agentic coding. GLM-5.2's differentiators are its 1 million token context, MIT license, and coding-benchmark leadership. Because scores shift with the test harness and model version, these rankings reflect a mid-2026 snapshot rather than a fixed ordering. ^[1]

How much does GLM-5.2 cost and where can you use it?

GLM-5.2 rolled out in stages: it reached Z.ai's GLM Coding Plan subscribers on June 13, 2026, the metered API opened on June 16, and the MIT open weights were published on Hugging Face on June 17, 2026. ^[5][3] Z.ai's first-party API is priced at $1.40 per million input tokens, $0.26 per million cached input tokens, and $4.40 per million output tokens, roughly one-sixth the cost of GPT-5.5. ^[1][6] Subscription GLM Coding Plans range from a Lite tier at a few dollars per month up to Max and Team tiers, and the open weights are additionally hosted by providers such as OpenRouter and DeepInfra, some of which list lower per-token rates. ^[5][7]

References

1. GLM-5.2 is the new leading open weights model on the Artificial Analysis Intelligence Index, Artificial Analysis, June 2026. ↩
2. GLM-5.2: Built for Long-Horizon Tasks, Z.ai / Zhipu AI (Hugging Face blog), June 17, 2026. ↩
3. zai-org/GLM-5.2 model card, Hugging Face. ↩
4. zai-org/GLM-5.2 config.json, Hugging Face. ↩
5. GLM 5.2 API and Pricing: GLM Coding Plan Guide, Lushbinary, June 2026. ↩
6. Z.ai's open-weights GLM-5.2 beats GPT-5.5 on multiple long-horizon coding benchmarks for 1/6th the cost, VentureBeat, June 2026. ↩
7. GLM 5.2 API pricing and benchmarks, OpenRouter. ↩
8. Zhipu AI releases harness for GLM-5.2 model as Chinese firm takes aim at Anthropic, South China Morning Post, June 2026. ↩