↑↓ to navigate Enter to select Esc to close
Try Pro
Setting things up for you...
Drop file here

Custom Instructions

Existing Custom Instructions Found

You have custom instructions from the previous version. Copy them to save before switching to the new format.

Custom instructions will only apply to new conversations. Start a new chat to see the changes in effect.

Your custom instructions are saved on-device only.

GPT-5 Settings

minimal low medium high
Controls depth of reasoning and problem-solving.

Notice: Reasoning effort setting only applies when using the Send button. Sending your message with the Enter key uses Send, fast reply (minimal effort) by default.
When disabled, Enter key sends messages according to reasoning effort setting
low medium high
Controls response detail and length

Claude Sonnet 4.5

We've upgraded the Claude Sonnet model from Sonnet 4 to Sonnet 4.5, Anthropic's most capable model to date. Sonnet now outperforms the latest Opus model (Claude Opus 4.1) while being significantly more cost-effective.

Capability
Anthropic's most capable model across most tasks
Context Window
200K tokens
Max Output
64K tokens
Knowledge Cutoff
January 2025
Cost
24x (vs Opus 120x)

Code execution and working with files: Claude can now run code, create downloadable files, and analyze data directly. File and image uploads are now available for Anthropic's models.

Artifacts
16

Not sure which one to pick?

For most tasks, we recommend Claude 4.5 Sonnet or GPT-5. Both are capable general-purpose models with different response styles.

Claude 4.5 Sonnet is Anthropic's most capable model, outperforming Claude Opus 4.1 on most tasks while being more cost-effective. Claude models support automatic code execution and can analyze documents, create downloadable files (PDFs, spreadsheets, charts), and work with uploaded images and files.

GPT-5 is a set of models with configurable reasoning and verbosity. Find these settings by selecting GPT-5 from the model selector, then clicking the settings icon. Click the lightning icon for quick responses with minimal reasoning.

For document uploads: We recommend Claude models (Sonnet 4.5 or Opus 4.1) or GPT-5. Both GPT-5 and the Claude models have excellent file-input support including visual understanding of PDFs. Claude models can also create downloadable outputs.

Claude Opus 4.1 is available for reference and specialized reasoning tasks, but Claude 4.5 Sonnet is more capable for most use cases.

o3 is a legacy reasoning model that performs well on math, science and coding. GPT-5 with high reasoning often surpasses o3 on benchmarks.

Gemini 2.5 Pro is Google's advanced multimodal model with 1M token context and code execution capabilities. Gemini 2.0 Flash is a faster alternative.

GPT-4o and GPT-4o mini are previous generation OpenAI models.

Moderate
Tech
Moderate
Creative

Share chat