> ## Documentation Index
> Fetch the complete documentation index at: https://docs.babbily.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Model capabilities
> Choose a Babbily model based on speed, reasoning, files, tools, media, context, and API usage budget.
Babbily gives you access to multiple models from one chat surface. The best model depends on what you are trying to do, how much context you need, and which capabilities are available for your plan.
Model availability, names, and capabilities can change over time. Use the model picker as the source of truth for what is available in your account.
## Choose by task
| If you need to... | Look for a model that is strong at... | Helpful setup |
| ------------------------------------ | -------------------------------------------- | -------------------------------------------------------------------------- |
| Draft, rewrite, or summarize quickly | Speed and clear writing | Keep tools **Off** unless the answer needs current information. |
| Reason through a complex plan | Deeper reasoning | Give goals, constraints, and decision criteria. |
| Analyze files | File support and enough context | Attach the most relevant files first. |
| Use current sources | Tool support | Use **Auto** or **Manual > Web search**. |
| Produce a broad research brief | Research support | Use **Manual > Deep research** and ask for citations. |
| Create or edit images | Image generation support | Describe subject, style, composition, lighting, and aspect ratio. |
| Create short videos | Video generation support | Describe motion, pacing, duration, and source media when supported. |
| Compare multiple approaches | Multi-perspective reasoning and synthesis | Ask for options, tradeoffs, and a final recommendation. |
| Work across a larger codebase | Long-context coding and repo-scale reasoning | Attach or paste the most relevant files and define the change boundary. |
| Work in a long thread | Larger context window | Watch the context indicator and start a new thread when the topic changes. |
| Save API usage budget | Lightweight model and fewer tools | Use model-only chat for simple drafts and rewrites. |
## Recently added models
Use the model picker as the source of truth for your account, but these recent additions can help when you are choosing quickly:
| Model | Best for | Notes |
| -------------- | ----------------------------------------------------------------------- | ---------------------------------------------------------------------------------- |
| GPT-5.6 Sol | Demanding long-horizon reasoning, coding, and complex agentic work | Choose Sol when capability matters more than speed or API usage budget. |
| GPT-5.6 Terra | Everyday reasoning, coding, and multi-step work | Choose Terra when you want a balance of strong capability and API usage budget. |
| GPT-5.6 Luna | Fast drafts, iteration, and lower-cost everyday work | Choose Luna when speed and API usage budget matter most within the GPT-5.6 family. |
| Grok 4.5 | Coding, tool use, STEM, and broad knowledge work | Useful for difficult tasks that require investigation, recovery, and verification. |
| Muse Spark 1.1 | Coding, debugging, long-running agentic tasks, and multimodal reasoning | Use it for extended workflows across supported text, document, and image inputs. |
## Capability signals
The model picker and composer can show what a model supports. Common capability differences include:
* File types and file size behavior.
* Context window size.
* Tool availability.
* Image generation or editing.
* Video generation or editing.
* Memory personalization availability.
* Speed and reasoning depth.
* Multi-step or agentic task behavior.
* Codebase-scale context handling.
* Fast mode availability.
* Sunsetting status for older model options.
* API usage budget impact.
If a file type, tool, or media action is hidden, the selected model may not support it.
When a model is marked **Sunsetting**, it remains available for continuity, but Babbily recommends choosing a newer model for new work when one is available.
## Advanced reasoning and coding work
Some models are built for specialized advanced work, such as comparing multiple approaches, reviewing code, working across large files, or planning multi-step engineering tasks. Choose these models when the task benefits from depth, synthesis, or a larger working context.
For better results:
* State the goal, constraints, and expected output.
* Include only the files or excerpts the model needs.
* Ask for a plan before large edits or refactors.
* Ask the model to call out uncertainty, risks, and tradeoffs.
## Switch models during work
You can switch models when the task changes.
Good switching patterns:
* Start with a faster model to brainstorm.
* Switch to a stronger reasoning model for decisions or plans.
* Switch to a file-capable model before attaching documents.
* Switch to a media-capable model for image or video work.
* Use Fast mode when it appears and you want quicker supported replies, then check your API usage budget for heavier work.
* Start a new thread if older context is steering the answer in the wrong direction.
When you switch models, restate the key goal if the thread is long. That helps the new reply focus on the current task.
## Usage considerations
More advanced model work can use more of your API usage budget, especially when it includes tools, long file context, image generation, video generation, or deep research.
For budget-sensitive work:
* Use shorter prompts when the task is simple.
* Avoid unnecessary file attachments.
* Turn tools **Off** for model-only drafting.
* Use search or deep research only when sources matter.
* Check your usage card before running image or video generation.
## Troubleshooting
If a capability is missing:
* Switch to a model that supports it.
* Change tool mode from **Off** to **Auto** or **Manual**.
* Check whether your plan supports the feature.
* Check your remaining API usage budget.
* Try a new thread if the current one has a long or confusing context history.
Learn the basics of choosing models and continuing threads.
Understand why capabilities can vary by model, plan, file, tool, and connector.