Zhipu · tested in Taskade Genesis
GLM 5.3 in Taskade
GLM 5.3 from Zhipu runs on the Max and Enterprise plans. We last tested it on Aug 29, 2026 in TSK-1, our app-building test.
- Status
- Available
- Plan
- Max and Enterprise
- Where to pick it
- More models or All Models
- Your own key
- No
- Last test
- Context window
- 1Mtokens
- Max output
- 131Ktokens
- Released
- Where it runs
- 19services
TSK-1 scorecard for the GLM family
TSK Score
GLM family, across real apps made in Taskade Genesis
- Interfaceready to share
- Emerging
- Taskcompletes your task
- Strong
- Memorykeeps your data
- Emerging
- Adapthandles follow-up edits
- Strong
Built a richer app than most, then ran out of time before finishing it.
MoreLess
On the long form it thought at length and never started. Both follow-up edits were clean. GLM 5.3 is now in the Taskade model picker, and this August test is its latest result.
Published scores
Benchmark scores that Zhipu and test sites published for GLM 5.3. Taskade did not re-run them. Lab-reported marks a score from Zhipu itself.
| Benchmark | Score | Source |
|---|---|---|
| Terminal-Benchpass@1 · v2.1 · max effort · Claude Code 2.1.207Hugging FaceLab-reported | 88.2% | Hugging FaceLab-reported |
| Humanity's Last Examaccuracy · max effort; with toolsHugging FaceLab-reported | 62.5% | Hugging FaceLab-reported |
More published scores (15)
| Benchmark | Score | Source |
|---|---|---|
| Terminal-Benchavg@3 · v3.0 · max effort · Claude Code 2.1.207Hugging FaceLab-reported | 28.3 | Hugging FaceLab-reported |
| DeepSWEresolved · v1.1 · max effort; 6h timeout; 400K context · mini-swe-agentHugging FaceLab-reported | 66.9% | Hugging FaceLab-reported |
| NL2Repomax effortHugging FaceLab-reported | 58 | Hugging FaceLab-reported |
| ProgramBenchalmost solved · max effortHugging FaceLab-reported | 19 | Hugging FaceLab-reported |
| FrontierSWEdominance score · max effort · Hugging FaceLab-reported | 78.1 | Hugging FaceLab-reported |
| SWE-Marathonv1.1 · max effort; modified anti-cheat checks · Claude Code 2.1.207Hugging FaceLab-reported | 42.5 | Hugging FaceLab-reported |
| PostTrainBenchweighted average · max effort; 3 runs; modified external-API checks · Claude Code 2.1.207Hugging FaceLab-reported | 39.8 | Hugging FaceLab-reported |
| CyberGympass@1 · max effort · Claude Code 2.1.207Hugging FaceLab-reported | 84.5% | Hugging FaceLab-reported |
| ExploitGymtasks solved · max effort; 2h rescaled timeout · Claude Code 2.1.207 · 869 tasksHugging FaceLab-reported | 105 | Hugging FaceLab-reported |
| ExploitGymtasks solved · max effort; 6h rescaled timeout · Claude Code 2.1.207 · 869 tasksHugging FaceLab-reported | 130 | Hugging FaceLab-reported |
| ExploitBenchaverage coverage · max effort; union over 3 revisions · Claude Code 2.1.207Hugging FaceLab-reported | 54.4 | Hugging FaceLab-reported |
| Toolathlon-Verifiedpass@1 · max effortHugging FaceLab-reported | 73% | Hugging FaceLab-reported |
| AutomationBenchpass@1 · v1.0.6 · max effortHugging FaceLab-reported | 48.2% | Hugging FaceLab-reported |
| Agents' Last Exammax effort; CLI · Claude Code 2.1.207Hugging FaceLab-reported | 28.5 | Hugging FaceLab-reported |
| GDPval-AAElo · v2 · max effortHugging FaceLab-reported | 1,769 | Hugging FaceLab-reported |
LMArena scores
| Benchmark | Score | Source |
|---|---|---|
| Text arena±5 · 22,543 votes · Rank 28 of 416 · LMArena | 1,479 | LMArena |
| WebDev arena±8 · 9,729 votes · Rank 23 of 145 · LMArena | 1,622 | LMArena |
Scores from the LMArena leaderboard dataset (CC BY 4.0). People vote between two answers, and the votes set the score. Taskade picked the Overall category and the default setting of each model.
Scores collected by models.dev (MIT). Each source links the page that published the score.
Versions tested
Open a version to read its findings.
GLM 5.3Tested Aug 20263 findings
Built a richer app than most, then ran out of time before finishing it. On the long form it thought at length and never started. Both follow-up edits were clean. GLM 5.3 is now in the Taskade model picker, and this August test is its latest result.
Four pages built, but the build did not finish inside the time limit.
On the 32-question client sign-up form it produced no app.
Both follow-up edits landed with no failed build actions.
Model facts
The token limits are the most the model accepts. Taskade can run it with lower limits.
- Last updated
- Input types
- Text
- Output types
- Text
- Reasoning
- Yes
- Tool calling
- Yes
- File attachments
- No
- Structured output
- Yes
- Open weights
- Yes
- Reasoning effort
- LowHighMax
Published specs from the model maker, via models.dev (MIT).
Where it runs
Services outside Taskade that serve GLM 5.3, from models.dev. Each link opens in a new tab.
How to choose GLM 5.3
- In the builder chat, open the model menu and choose More models. Then pick the model.
- In the agent editor, pick the model under All Models.
- If your model menu does not show the model, Taskade uses TSK-Auto instead.
Where GLM ranks
TSK Score out of 100.
Compare GLM
Other Zhipu models
FAQ
Which Taskade plans include GLM 5.3?
GLM 5.3 runs in Taskade on the Max and Enterprise plans. If your model menu does not show it, Taskade uses TSK-Auto instead.
Has Taskade tested GLM 5.3?
Yes. We last tested GLM 5.3 on Aug 29, 2026 in the TSK-1 Benchmark. Built a richer app than most, then ran out of time before finishing it. On the long form it thought at length and never started. Both follow-up edits were clean. GLM 5.3 is now in the Taskade model picker, and this August test is its latest result.
How do I choose GLM 5.3 in Taskade?
In the builder chat, open the model menu and choose More models. Then pick GLM 5.3. In the agent editor, pick it under All Models.
Who makes GLM 5.3?
Zhipu makes GLM 5.3. Taskade runs it on the Max and Enterprise plans, beside frontier models from other top AI labs.