GLM-5.1
- Made by
- Z.ai
- Released
- Announced in
- GLM-5.1: Towards Long-Horizon Tasksz.ai
- Event
- GLM-5.1
- API record
- /api/v1/models/glm-5-1
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| Humanity's Last ExamNo tools · text-only subset; GPT-5.2 judge; up to 163,840 output tokens | 31.0% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| Humanity's Last ExamWith tools · text-only subset; GPT-5.2 judge; 202,752-token context | 52.3% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.CorroboratedThe same number is printed in 1 other lab document: 1 |
| SWE-bench ProOpenHands with a tailored prompt; 200K context · The source doesn't say whether this is the public or the commercial set | 58.4% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| GPQA Diamondup to 163,840 output tokens | 86.2% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| BrowseCompwithout context management: keeps the last 5 turns | 68.0% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| with context management, discard-all strategy | 79.3% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Terminal-Bench 2.0Terminus 2 harness; 3h timeout; 200K context | 63.5% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Claude Code harness, thinking; no time limit; averaged over 5 runs | 69.0% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| DeepSWE v1.1mini-swe-agent harness · reported later, in GLM-5.2: Built for Long-Horizon Tasks | 18.0% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Terminal-Bench 2.1Terminus 2 harness · reported later, in GLM-5.2: Built for Long-Horizon Tasks | 63.5% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| Claude Code harness | 69% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| MCP Atlas500-task public set; thinking mode; 10-minute timeout per task; Gemini 3 Pro judge | 71.8% | CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
Sources
Reported by Z.ai
- GLM-5.1: Towards Long-Horizon Tasksz.ai · 7 Apr 2026
Our capture of the source - GLM-5.2: Built for Long-Horizon Tasksz.ai · 16 Jun 2026
Our capture of the source