Meta Muse vs. ChatGPT, Claude and Grok: Who can do what?

Meta has launched Muse Spark 1.1, the latest version of its AI model. It has now been compared with the three big competitors: OpenAI's ChatGPT, Anthropic's Claude and xAI's Grok. The figures below come from a comparison on the Eden AI site and are benchmarks, meaning standardized tests. They do not necessarily say anything about how the models feel in everyday use.

Four models, four strengths

GPT-5.6 Sol from OpenAI is best at pure coding. It leads on the Terminal-Bench test, which measures how well an AI can solve tasks in a command-line environment.

Claude Fable 5 gets the highest overall score for general intelligence: 60 points on the Artificial Analysis index.

Grok 4.5 is best when you look at price relative to performance. It costs $2 for input and $6 for output per million tokens and has a context window of 500,000 tokens. A token is a small piece of text, and a context window is how much text the model can hold in its head at once.

Muse Spark 1.1 is best at controlling tools. This is called agentic use: the AI plans several steps, uses programs and carries out actions across apps with limited help from humans. On the other hand, it lags behind on pure coding tests such as SWE-Bench Pro.

What can Muse Spark do?

The model understands text, images and video but responds only in text. It has a very large context window of about one million tokens. Meta highlights, among other things, a shopping mode that uses what you already follow to suggest clothes and products. It can also build websites and small games from a description and answer common health questions, such as calories from a photo of food.

Muse Spark powers the Meta AI app and website. According to coverage, it is rolling out in WhatsApp, Instagram, Facebook and Messenger as well as Meta's AI glasses.

Price

Muse Spark 1.1 is Meta's first paid AI model that developers can use via API. The price is $1.25 per million input tokens and $4.25 for output. That is significantly cheaper than Claude Fable 5, where the input price according to the comparison is about eight times higher.

What does it mean for you?

If you use Facebook, Instagram or WhatsApp, you will probably encounter Muse without choosing it. For writing, coding or research, you should choose based on the task. No model wins at everything.

What it means for you

If you have Facebook, Instagram or WhatsApp, you may soon come across Meta's new AI without choosing it yourself, and you can already try it in the Meta AI app or on the website. Use it, for example, to make a simple website or a small game from a description, or to get an estimate of the calories in a photo of your food. If you need help with writing, coding or research, choose based on the task and feel free to try two models on the same question, because none of them wins at everything.