| Benchmark | GLM-5.3 | GLM-5.2 | Kimi K3 | DeepSeek-V4 Pro-0813 | Qwen3.8-Max | Opus 4.8 | Fable 5 (w/ fallback) | GPT-5.6 Sol |
| ExploitGym (2h / 6h) | 105 / 130 | 29 / 39 | 36 / 70 | – | 14 / 26 | 80 / 120 | 181 / 247 | **216 / 293** |
- **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating.
GLM-5.3 supports controlling the thinking budget through the `reasoning_effort` parameter, which accepts three levels: `low`, `high`, and `max`. It defaults to `max` if not passed (or if set to any other value). To use `low` or `high`, pass them explicitly. For benchmark and leaderboard reproduction, keep the default `max`.
The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating.
条件
GLM-5.3目标列;The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating. 安全、内部子项或跨测试聚合仅保留原始披露,不进入常规展示。
GLM-5.3 · ExploitGym (2h / 6h) (GLM-5.3模型卡;2h / 6h) 105 / 130 · Main text evidence (ev-gmc-glm-5-3-84)
Official source for ExploitGym (2h / 6h) · GLM-5.3模型卡;2h / 6h in GLM-5.3. GLM-5.3 · max · 105 / 130. See “GLM-5.3 official model card” at Benchmark table;ExploitGym (2h / 6h);GLM-5.3目标列. Evidence ev-gmc-glm-5-3-84.
GLM-5.3 · ExploitGym (2h / 6h) (GLM-5.3模型卡;2h / 6h), configuration GLM-5.3 · max, pass-at-1 105 / 130 (exact score publication date unconfirmed). See “GLM-5.3 official model card” at Benchmark table;ExploitGym (2h / 6h);GLM-5.3目标列. View score conditions
| Benchmark | GLM-5.3 | GLM-5.2 | Kimi K3 | DeepSeek-V4 Pro-0813 | Qwen3.8-Max | Opus 4.8 | Fable 5 (w/ fallback) | GPT-5.6 Sol |
| ExploitGym (2h / 6h) | 105 / 130 | 29 / 39 | 36 / 70 | – | 14 / 26 | 80 / 120 | 181 / 247 | **216 / 293** |
- **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating.
GLM-5.3 supports controlling the thinking budget through the `reasoning_effort` parameter, which accepts three levels: `low`, `high`, and `max`. It defaults to `max` if not passed (or if set to any other value). To use `low` or `high`, pass them explicitly. For benchmark and leaderboard reproduction, keep the default `max`.
The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating.
Conditions
GLM-5.3目标列;The card says reasoning_effort defaults to max and recommends keeping that default for benchmark reproduction. **ExploitGym**: We evaluate GLM-5.3, Kimi-K3 and Qwen3.8 Max in Claude Code 2.1.207 (max reasoning effort, no web tools with `temperature=1.0`, `top_p=1.0`, `max_new_tokens=128000`). The reported results are single-run Pass@1 on 869 tasks under two timeout budgets: 2 hours and 6 hours, which are calculated as the API inference time rescaled by per-model tokens per second rate (per-model TPS sourced from Artificial Analysis; that is, we rescale GLM-5.3's results by 115 TPS, Kimi K3's results by 40 TPS and Qwen3.8 Max's results by 47 TPS), plus the non-API overhead. We also apply a domain whitelist (allowing only essential domains such as pypi.org and deb.debian.org for basic tool installation) to prevent the agent from cheating. 安全、内部子项或跨测试聚合仅保留原始披露,不进入常规展示。