Models
Every model in the Hyper catalog is hand-picked for software development. We only include coding models with reasoning capabilities, optimized for long running coding sessions in interactive workflows and fully agentic software development lifecycle workflows. You won't find general-purpose chat models here. Every model on the list has been vetted to perform well at code generation, debugging, refactoring, and multi-file editing tasks.
Key Fields
| Field | Description |
|---|---|
| Context | Maximum number of tokens the model can process in a single request. Larger context means you can feed it more code, docs, or conversation history at once. |
| Input | Cost per 1M tokens sent to the model. This covers your prompts, system instructions, and any attached context. |
| Output | Cost per 1M tokens generated by the model. This covers completions, code suggestions, and reasoning traces. |
| Cache Read | Cost per 1M tokens read from the prompt cache. Cheaper than full input pricing because the model provider already has them in memory. |
| Cache Write | Cost per 1M tokens written to the prompt cache. You pay this once to warm the cache, then subsequent reads are cheaper. |
Pricing
We always aim to drive the most cost-effective pricing across inputs, outputs, and cache hits to ensure optimal hypercredit consumption. Hypercredits are computed based on total token cost with 1 hypercredit equating to $0.05.
Features
Models may have one or more feature badges:
| Badge | Description |
|---|---|
| Reasoning | The model supports extended thinking or chain-of-thought. Useful for complex architecture decisions and multi-step debugging. |
| Attachments | The model accepts image and document uploads alongside text. Handy for screenshots, diagrams, and log files. |
Attachments in Practice
Image attachments are especially useful in software development scenarios such as UI development, where you can share screenshots of a broken layout or a design mockup directly with the model. Visual bug reports, diagram-to-code workflows, and design review all benefit from sending images alongside your prompts.
Models that currently support attachments in the Hyper catalog include DeepSeek V4 Flash and GLM 5.1. Check the model card for the attachment badge to confirm support before sending images.
Unsupported Attachments
Hyper returns an error when sending attachments to models that do not support them. Always verify that a model has the Attachments badge before including images or documents in your request.