GLM-5.3 is the flagship open-source AI model — the most capable open-weights model for coding, with a stable context of one million tokens. It reports 88.2 on Terminal-Bench 2.1, 78.1 on FrontierSWE, and a state-of-the-art 84.5 on CyberGym for vulnerability discovery. Below, you can compare each model — or experience the free AI sandbox first.
No registration required, no account required, no quota required. Real-time streaming responses can be obtained by entering prompt words.
The reply will be streamed out here...
This experience uses the GLM large language model, and real-time streaming responses can be obtained by entering prompt words.
Four capabilities define the current generation of models — all reported by the model authors and reproducible from open rights.
A context of one million tokens is not only available but also maintains quality: GLM-5.3 carries over training on long-context coding agent trajectories such as large-scale implementation, automated research, and complex debugging.
State of the art on CyberGym (84.5) for vulnerability discovery. In real-world testing it identified 2,436 vulnerabilities across 269 open-source projects — the oldest dating back to 1981.
Every four sparse attention layers share a lightweight indexer, reducing the FLOPs per token by 2.9 times under a context length of one million, while maintaining long-distance quality without loss.
Open-source models under the MIT license: weights are published on HuggingFace and ModelScope, without regional restrictions, and support local deployment through transformers, vLLM, SGLang, xLLM, and ktransformers.
All the following data are the results of GLM-5.3 and comparative models published by the model authors. glmmodel.net only provides reports and does not run these evaluations.
Task duration can reach 20 hours
Up to 10 hours on a single H100 card
Ultra-long cycle software engineering, up to 10 hours
Across these long-horizon benchmarks, GLM-5.3 is the highest-ranking open-source model — proof that its million-token context converts into real output. It improves from 4.6 to 28.3 on Terminal-Bench 3.0, from 46.2 to 66.9 on DeepSWE v1.1, and more than doubles GLM-5.2 on exploitation benchmarks — every gain coming from post-training on the same base model.
Published GLMMODEL benchmark results, covering 16 evaluations of 8 models.
| Benchmark Test | GLM-5.3 | GLM-5.2 | Kimi K3 | DeepSeek-V4 Pro | Qwen3.8-Max | Claude Opus 4.8 | Fable 5* | GPT-5.6 Sol |
|---|---|---|---|---|---|---|---|---|
| Coding | ||||||||
| Terminal-Bench 2.1 | 88.2 | 81.0 | 88.3 | 87.9 | 86.6 | 85.0 | 88.0 | 88.8 |
| Terminal-Bench 3.0 | 28.3 | 4.6 | 17.4 | – | – | 21.1 | 33.7 | 34.6 |
| DeepSWE v1.1 | 66.9 | 46.2 | 67.5 | 62.7 | 56.6 | 58.0 | 69.7 | 72.7 |
| NL2Repo | 58.0 | 48.9 | 58.0 | 61.1 | 55.9 | 69.7 | – | – |
| ProgramBench (Almost Solved) | 19.0 | 9.5 | 17.5 | – | 10.5 | 15.5 | 33.0 | 23.0 |
| FrontierSWE | 78.1 | 67.5 | – | – | – | 66.5 | 88.2 | – |
| SWE-Marathon v1.1 | 42.5 | 19.4 | 48.1 | – | – | 48.8 | 33.1 | 42.5 |
| PostTrainBench | 39.8 | 31.7 | 32.0 | – | – | 32.9 | 41.8 | 36.2 |
| Cyber | ||||||||
| CyberGym | 84.5 | 77.2 | 80.0 | 83.3 | 78.5 | 78.1 | 83.8 | 83.6 |
| ExploitGym (2h / 6h) | 105 / 130 | 29 / 39 | 36 / 70 | – | 14 / 26 | 80 / 120 | 181 / 247 | 216 / 293 |
| ExploitBench | 54.4 | 24.4 | 32.2 | – | 28.8 | 40.0 | 78.0 | 76.5 |
| Agent | ||||||||
| Toolathlon Verified | 73.0 | 59.9 | 76.5 | 74.1 | 72.5 | 76.2 | 74.7 | 74.9 |
| AutomationBench v1.0.6 | 48.2 | 26.2 | 46.7 | 43.2 | 39.8 | 41.0 | 46.2 | 45.8 |
| Agents' Last Exam (ALE-CLI) | 28.5 | 23.8 | 27.6 | 25.7 | 27.0 | 25.7 | 23.8 | 28.6 |
| HLE (With Tools) | 62.5 | 54.7 | 59.8 | 60.0 | 56.2 | 57.9 | 63.9 | 64.5 |
| GDPval-AA v2 | 1769 | 1508 | 1682 | 1590 | 1739 | 1588 | 1743 | 1730 |
* Fable 5 results are reported with fallback. All data are results published by the model authors.
GLM-5.3 always thinks, at three intensity levels. Effort control lets you spend compute only when the task demands it — GLM-5.3 beats GLM-5.2 at every effort level while consuming fewer output tokens.
lightweight reasoning
Scenarios where latency matters more than depth, such as quick editing, template code, and routine refactoring — GLM-5.3 no longer disables thinking, so light reasoning stays on by default.
~50K average tokens
The default for most agent coding: at High effort GLM-5.3 reaches 31.4% with about 50K output tokens, surpassing Claude Opus 4.8 (29.5% at 120K).
~75K average tokens
The hardest long-horizon problems: 34.5% at roughly 75K output tokens, versus GLM-5.2's 23.4% at 96K — more results with fewer tokens.
Data from Z.ai Code Bench v1.0, an in-house benchmark that evaluates coding agents in realistic local development environments, measuring end-to-end task completion at controlled effort levels. Reported by the model authors. GLM-5.3 improves both performance and token efficiency over GLM-5.2 at every effort level.
Increasing the context limit from 200K to one million tokens is an engineering issue rather than a configuration parameter. Three modifications are key.
Every four Transformer layers share a lightweight indexer. It is located at the first layer among the four, and its top-k index is reused by the other three layers — eliminating the dot product and top-k calculation in three-quarters of the layers. Result: Each token's FLOPs is reduced by 2.9 times under a context length of one million. IndexShare was introduced starting from a sequence length of 128K during the training mid-phase.
The multi-token prediction layer runs the indexer once in the first draft step and reuses it in subsequent steps, eliminating the mismatch between the training/inference KV cache of the previous generation. Combined with rejection sampling and end-to-end TV loss, the accepted length increases from 4.56 to 5.47 — about 20% — across 7 MTP steps.
| Baseline | 4.56 |
| + IndexShare + KVShare | 5.10 |
| + Reject Sampling | 5.29 |
| + End-to-End TV Loss | 5.47 (+20%) |
After more than a few tens of thousands of tokens, the bottleneck is no longer computation, but the KV cache capacity, long context kernel, and CPU overhead. Three fixes: fine-grained memory management and parallelization based on LayerSplit, kernel optimization for context length scaling coordinated with the cache transfer pipeline, and CPU-side cache management, request scheduling, and runtime path optimization. The throughput advantage expands with the growth of context.
Normalized Throughput Advantage
The training stack (slime) combines white-box and black-box rollouts, compact trajectories, and sub-agent workflows. Parallel OPD training merges more than a dozen expert models in about two days, and FP8 KV caching keeps rollout memory within budget. Through GLM-5.3, system-level optimizations improved end-to-end RL training throughput by more than 2.3×.
Long-term RL can induce shortcuts, so suspicious actions are first passed through a rule-based recall filter, and then checked for accuracy by the LLM judge. Confirmed malicious actions are intercepted online and responded with virtual tool results, allowing rollouts to continue rather than be discarded. Based on Critic's PPO, compressible sub-traces are maintained for training on a single rollout.
From 30B Flash level to a million-context flagship — choose a model to view the complete specification table and benchmark data.
The most capable open-weights model for coding: +50% on Z.ai Code Bench, open-source SOTA on Terminal-Bench 3.0 and Agents' Last Exam, state of the art on CyberGym.
First natively multimodal GLM-5 model: 320B total / 18B active parameters, hybrid attention, 1M context — GLM-5.2-beating performance at one-tenth the price.
Previous-generation flagship: one million token context and the long-horizon foundation that GLM-5.3's post-training builds on.
Long-horizon specialist: significantly improved long-term capabilities and engineering-level outputs that can work autonomously for several hours.
Stronger programming capabilities, more reliable multi-step execution, and better complex agent behavior.
Basic model, optimized for long chain and dynamic intelligent agent scenarios.
Enhanced programming capabilities, stable multi-step inference, and improved frontend generation.
300 billion parameters; high efficiency and performance, leading among models of the same scale.
Multi-file functionality beyond a single context window: the model keeps the entire repository state in view, rather than re-reading it every few rounds.
Several hours of experimental cycles, agents need to plan, run, read results, and correct — this is exactly the workload designed to be evaluated by FrontierSWE and PostTrainBench.
Complete call graphs, analyzer outputs, and keeping the kernel and system work on the same track as previous attempts are required.
Long-term reproducibility tracking, unstable testing, and cross-service failures; truncating the context is the reason for losing bugs.
All the above scenarios can be run on your own hardware — MIT weights, transformers · vLLM · SGLang · xLLM · ktransformers.How to run models locally →
Read the GLMMODEL benchmark tests, and then let the real model start working. The sandbox above is free and requires no information; if you want a more complete AI toolkit, the free package from the partners starts here.