We are launching GLM-5, targeting complex systems engineering and long-horizon agentic tasks. Scaling is still one of the most important ways to improve the intelligence efficiency of Artificial General Intelligence (AGI). Compared to GLM-4.5, GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active), and increases pre-training data from 23T to 28.5T tokens. GLM-5 also integrates DeepSeek Sparse Attention (DSA), largely reducing deployment cost while preserving long-context capacity.
Reinforcement learning aims to bridge the gap between competence and excellence in pre-trained models. However, deploying it at scale for LLMs is a challenge due to the RL training inefficiency. To this end, we developed slime, a novel asynchronous RL infrastructure that substantially improves training throughput and efficiency, enabling more fine-grained post-training iterations. With advances in both pre-training and post-training, GLM-5 delivers significant improvement compared to GLM-4.7 across a wide range of academic benchmarks and achieves best-in-class performance among all open-source models in the world on reasoning, coding, and agentic tasks, closing the gap with frontier models.
Benchmark
Table with columns: GLM-5, GLM-4.7, DeepSeek-V3.2, Kimi K2.5, Claude Opus 4.5, Gemini 3 Pro, GPT-5.2 (xhigh)
GLM-5
GLM-4.7
DeepSeek-V3.2
Kimi K2.5
Claude Opus 4.5
Gemini 3 Pro
GPT-5.2 (xhigh)
HLE
30.5
24.8
25.1
31.5
28.4
37.2
35.4
HLE (w/ Tools)
50.4
42.8
40.8
51.8
43.4*
45.8*
45.5*
AIME 2026 I
92.7
92.9
92.7
92.5
93.3
90.6
*: refers to their scores of full set.
†: A verified version of Terminal-Bench 2.0 that fixes some ambiguous instructions.
See footnote for more evaluation details.
Footnote
Humanity’s Last Exam (HLE) & other reasoning tasks: We evaluate with a maximum generation length of 131,072 tokens (temperature=1.0, top_p=0.95, max_new_tokens=131072). By default, we report the text-only subset; results marked with * are from the full set. We use GPT-5.2 (medium) as the judge model. For HLE-with-tools, we use a maximum context length of 202,752 tokens.
SWE-bench & SWE-bench Multilingual: We run the SWE-bench suite with OpenHands using a tailored instruction prompt. Settings: temperature=0.7, top_p=0.95, max_new_tokens=16384, with a 200K context window.
BrowserComp: Without context management, we retain details from the most recent 5 turns. With context management, we use the same discard-all strategy as DeepSeek-v3.2 and Kimi K2.5.
Terminal-Bench 2.0 (Terminus 2): We evaluate with the Terminus framework using timeout=2h, temperature=0.7, top_p=1.0, max_new_tokens=8192, with a 128K context window. Resource limits are capped at 16 CPUs and 32 GB RAM.
Terminal-Bench 2.0 (Claude Code): We evaluate in Claude Code 2.1.14 (think mode, default effort) with temperature=1.0, top_p=0.95, max_new_tokens=65536. We remove wall-clock time limits due to generation speed, while preserving per-task CPU and memory constraints. Scores are averaged over 5 runs. We fix environment issues introduced by Claude Code and also report results on a verified Terminal-Bench 2.0 dataset that resolves ambiguous instructions (see: https://huggingface.co/datasets/zai-org/terminal-bench-2-verified).
Serve GLM-5 Locally
Prepare environment
vLLM, SGLang, KTransformers, and xLLM all support local deployment of GLM-5. A simple deployment guide is provided here.
CyberGym: We evaluate in Claude Code 2.1.18 (think mode, no web tools) with (temperature=1.0, top_p=1.0, max_new_tokens=32000) and a 250-minute timeout per task. Results are single-run Pass@1 over 1,507 tasks.
MCP-Atlas: All models are evaluated in think mode on the 500-task public subset with a 10-minute timeout per task. We use Gemini 3 Pro as the judge model.
τ²-bench: We add a small prompt adjustment in Retail and Telecom to avoid failures caused by premature user termination. For Airline, we apply the domain fixes proposed in the Claude Opus 4.5 system card.
Vending Bench 2: Runs are conducted independently by Andon Labs.