Zhipu AI releases GLM-5.2 with 1M token context and 62.1% SWE-bench Pro score
Zhipu AI released GLM-5.2, a 753 billion parameter model with a 1 million token context window. The model scores 62.1% on SWE-bench Pro and introduces IndexShare architecture that reduces per-token FLOPs by 2.9× at 1M context length. Released under MIT license with no regional restrictions.
GLM-5.2 — Quick Specs
Zhipu AI releases GLM-5.2 with 1M token context and 62.1% SWE-bench Pro score
Zhipu AI released GLM-5.2, a 753 billion parameter model with a 1 million token context window and enhanced coding capabilities. The model scores 62.1% on SWE-bench Pro, positioning it between GPT-5.5 (58.6%) and Claude Opus 4.8 (69.2%) according to the company's benchmarks.
Key specifications
GLM-5.2 achieves the following benchmark scores according to Zhipu AI:
- SWE-bench Pro: 62.1%
- NL2Repo: 48.9%
- DeepSWE: 46.2%
- ProgramBench: 63.7%
- AIME 2026: 99.2%
- GPQA-Diamond: 91.2%
- HMMT November 2025: 94.4%
The model supports 1 million token context length and is available in FP8 quantization format. Parameter count is 753 billion. Pricing has not been disclosed.
Technical architecture
GLM-5.2 introduces IndexShare, which reuses the same indexer across every four sparse attention layers. According to Zhipu AI, this reduces per-token FLOPs by 2.9× at 1M context length compared to the previous architecture.
The model includes an improved MTP (Multi-Token Prediction) layer for speculative decoding, which the company claims increases acceptance length by up to 20%.
GLM-5.2 offers multiple "thinking effort levels" for coding tasks, allowing developers to balance between performance and latency depending on task complexity.
Deployment and availability
The model is released under MIT open-source license with no regional restrictions. It supports deployment through:
- SGLang (v0.5.13.post1+)
- vLLM (v0.23.0+)
- Transformers (v0.5.12+)
- KTransformers (v0.5.12+)
For Ascend NPU platforms, vLLM-Ascend, xLLM, and SGLang are supported.
API access is available through Z.ai API Platform, though pricing details have not been published.
Benchmark positioning
On coding benchmarks, GLM-5.2 shows improvements over its predecessor GLM-5.1 (58.4% on SWE-bench Pro) and claims to match or exceed Qwen3.7-Max (60.6%) and MiniMax M3 (59%). However, it trails Claude Opus 4.8 (69.2%) and DeepSeek-V4-Pro (55.4%) on several metrics.
On the HLE reasoning benchmark, GLM-5.2 scores 40.5, behind Claude Opus 4.8 (49.8) and Gemini 3.1 Pro (45), but ahead of DeepSeek-V4-Pro (37.7) and MiniMax M3 (37).
What this means
GLM-5.2 represents Zhipu AI's push into the 1M context space dominated by models like Gemini 2.5 Pro and Claude 3.7 Sonnet. The IndexShare architecture addresses a key challenge in long-context models: computational efficiency at extended lengths. The 2.9× reduction in FLOPs at 1M tokens could make the model more practical for deployment compared to architectures that scale linearly with context length. The MIT license removes barriers common in Chinese AI models, potentially increasing adoption in Western markets. However, without disclosed pricing and independent benchmark verification, actual competitiveness against Anthropic and OpenAI's offerings remains to be demonstrated in production environments.
Related Articles
LG AI Research Releases K-EXAONE 2.0, a 750B-Parameter Open-Weight MoE Model with 262K Context
LG AI Research has released K-EXAONE 2.0, a 750-billion-parameter mixture-of-experts language model with 37B active parameters, a 262,144-token context window, and support for 10 languages. The model is open-weighted under Apache 2.0 and claims competitive results against Qwen3.5, GLM-5.1, and DeepSeek-V4 Pro on reasoning, coding, and long-context benchmarks.
OpenAI Halts Parts of Astra Model Development After It Hit 'Critical' Cybersecurity Threshold
OpenAI disclosed that its in-development Astra model showed cyberattack capabilities strong enough that it cannot rule out a 'Critical' risk classification. The company has paused related internal activity and added security controls under its Preparedness Framework.
Mistral's 3B-Parameter Shieldstral Matches 20B Safety Model on Text Benchmarks
Mistral's new Shieldstral, a 3-billion-parameter open-weight safety classifier, posts an 84.9% F1 score on text benchmarks—tying OpenAI's GPT-OSS-Safeguard-20B, a model roughly seven times larger. The model lets operators define safety rules at runtime using plain-language yes/no questions instead of fixed taxonomies.
Mistral AI Releases Shieldstral-1.0-3B, a 3B-Parameter Policy-Adaptive Safety Classifier
Mistral AI has released Shieldstral-1.0-3B, a compact open-weight safety classifier that evaluates text and images against natural-language policies specified at inference time. The 3B model runs on a single GPU and reports F1 scores competitive with or exceeding larger moderation models like LlamaGuard-4-12B and GPT-OSS-Safeguard-20B on multiple benchmarks.
Comments
Loading...