OpenAI · tested in Taskade Genesis
GPT-5.6 Luna in Taskade
GPT-5.6 Luna from OpenAI runs on every paid plan. We last tested it on Oct 8, 2026 in TSK-1, our app-building test.
- Status
- Available
- Plan
- Every paid plan
- Where to pick it
- More models or All Models
- Your own key
- OpenAI
- Last test
- Context window
- 1.05Mtokens
- Max output
- 128Ktokens
- Released
- Knowledge cutoff
- Where it runs
- 8services
TSK-1 scorecard for the GPT family
TSK Score
GPT family, across real apps made in Taskade Genesis
- Interfaceready to share
- Strong
- Taskcompletes your task
- Leading
- Memorykeeps your data
- Strong
- Adapthandles follow-up edits
- Leading
The most faithful model we have tested.
MoreLess
It reproduced all 32 questions of a real customer's client sign-up form word for word again, built a scoring automation that ran, and passed every September follow-up-edit test.
Published scores
Benchmark scores that OpenAI and test sites published for GPT-5.6 Luna. Taskade did not re-run them. Lab-reported marks a score from OpenAI itself.
| Benchmark | Score | Source |
|---|---|---|
| SWE-Bench Proresolve rate · OpenAILab-reported | 62.7% | OpenAILab-reported |
| Terminal-Benchsuccess rate · v2.1 · OpenAILab-reported | 84.7% | OpenAILab-reported |
| OSWorldsuccess rate · v2.0 · OpenAILab-reported | 45.6% | OpenAILab-reported |
| BrowseCompaccuracy · OpenAILab-reported | 83.3% | OpenAILab-reported |
| Toolathlonsuccess rate · OpenAILab-reported | 53.4% | OpenAILab-reported |
| GPQA Diamondaccuracy · OpenAILab-reported | 92.3% | OpenAILab-reported |
| FrontierMathaccuracy · v2 · Tier 1-3 · OpenAILab-reported | 78.6% | OpenAILab-reported |
| MMMU Proaccuracy · no tools · OpenAILab-reported | 78.4% | OpenAILab-reported |
More published scores (4)
| Benchmark | Score | Source |
|---|---|---|
| DeepSWEresolve rate · v1.1 · OpenAILab-reported | 67.2% | OpenAILab-reported |
| Agents' Last ExamOpenAILab-reported | 50.3 | OpenAILab-reported |
| Artificial Analysis Intelligence Indexindex score · v4.1 · max · Artificial Analysis | 51.2 | Artificial Analysis |
| Artificial Analysis Coding Agent Indexindex score · v1.1 · max · Codex · Artificial Analysis | 74.6 | Artificial Analysis |
Scores collected by models.dev (MIT). Each source links the page that published the score.
Versions tested
Open a version to read its findings.
GPT-5.6 LunaTested Oct 202620 findings
The most faithful model we have tested. It reproduced all 32 questions of a real customer's client sign-up form word for word again, built a scoring automation that ran, and passed every September follow-up-edit test.
Reproduced all 32 questions of a real customer's client sign-up form word for word. The only model to get all four of its builds word for word across more than one test.
The only model to build a working scoring grid: 33 rows by 5 ratings, with every score saved straight back into the workspace.
The best client sign-up form Luna has built: nothing stalled, and a human reviewer called it beautiful.
Test winner on value. The only model to explain its own design choices, and both builds carried all 32 questions word for word.
The baseline for automatic model picking: an app built in about 11 minutes and edited in about 6.
Its match tracker crashed on the first real entry and its sign-up form rejected the first submission; later tests finished cleanly.
The value pick on every app in a five-model test of identical prompts.
Unchanged result after a large platform update, with fewer failed build actions.
Kept a build checklist inside both apps, so the finished work could be checked against the request.
Built the client sign-up form with four pages and a scoring automation that ran on its own.
Built the fleet-inspection app with both weekly automations live in 7 minutes 18 seconds.
Built the AI governance office in 6 minutes 38 seconds with all 24 named items in place.
Built a complete match tracker with a live dashboard and a weekly recap automation.
Passed all five follow-up-edit tests on its habit tracker.
Its recipe box rendered as a warm printed cookbook in light and dark, with every page loading cleanly.
Built a two-page match tracker in 8 minutes 19 seconds, the fastest and most economical complete build of its test, and a match saved from the app came back with all eight details in place.
Built the family intake form so it saved every answer and both photos, and added the new teacher question cleanly: the most economical complete build of its test.
Its public class assistant told an anonymous visitor a child's name from a saved form.
Built the family intake form twice. One build saved every answer and both photos in under 5 minutes. In the other, every parent submission failed on the photo permission question, although its closing summary said the form had worked.
Built the Dota 2 match tracker twice in about 5 and a half minutes each. Both saved a logged match with every detail and added a ranked heroes page when asked.
Results by task
Model facts
The token limits are the most the model accepts. Taskade can run it with lower limits.
- Last updated
- Input types
- Text, Images, PDF
- Output types
- Text
- Reasoning
- Yes
- Tool calling
- Yes
- File attachments
- Yes
- Structured output
- Yes
- Open weights
- No
- Reasoning effort
- NoneLowMediumHighExtra highMax
Published specs from the model maker, via models.dev (MIT).
Run GPT-5.6 Luna with your own key
Taskade runs it for you. You can also pick it with your own key in a custom agent.
| Key provider | Plan |
|---|---|
| OpenAI | Pro and up |
Where it runs
Services outside Taskade that serve GPT-5.6 Luna, from models.dev. Each link opens in a new tab. In Taskade, your own key works only through the key providers on this page.
How to choose GPT-5.6 Luna
- 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 GPT ranks
TSK Score out of 100.
Compare GPT
Other OpenAI models
FAQ
Which Taskade plans include GPT-5.6 Luna?
GPT-5.6 Luna runs in Taskade on every paid plan. If your model menu does not show it, Taskade uses TSK-Auto instead.
Has Taskade tested GPT-5.6 Luna?
Yes. We last tested GPT-5.6 Luna on Oct 8, 2026 in the TSK-1 Benchmark. The most faithful model we have tested. It reproduced all 32 questions of a real customer's client sign-up form word for word again, built a scoring automation that ran, and passed every September follow-up-edit test.
How do I choose GPT-5.6 Luna in Taskade?
In the builder chat, open the model menu and choose More models. Then pick GPT-5.6 Luna. In the agent editor, pick it under All Models.
Can I use my own API key for GPT-5.6 Luna?
Yes. In a custom agent, you can pick GPT-5.6 Luna with your own key from OpenAI (Pro and up). Your provider bills the model use to your own account.
Who makes GPT-5.6 Luna?
OpenAI makes GPT-5.6 Luna. Taskade runs it on every paid plan, beside frontier models from other top AI labs.