Compare Fable 5.2 and Gemini 4 Carbon with one prompt - pick the best model for your task in 5 minutes
Run the same test request in both models and determine the winner by three criteria: quality, speed, price
- 1.Open Google AI Studio and select the gemini-4-carbon model from the dropdown list. If the model is not in the list, the paid API access has not yet been rolled out in your region; try again later or use an alternative access gateway.
Carbon is only available on the paid API tier. There is no free plan for this model.
Google AI Studio ↗ - 2.Paste the following prompt into the chat, replacing [DESCRIBE YOUR ACTUAL TASK HERE] with a real description of the task you want to assign to the AI - for example, «write a Python script that reads a CSV file and generates a summary report» or «analyze this contract text and extract three key risks: [insert text]».
The same prompt goes into both models - do not change the wording between runs, otherwise the comparison will be incorrect.
Google AI Studio ↗PromptSolve this task and format your response in three sections: 1. SOLUTION: your complete answer with all necessary code, text, or analysis 2. ASSUMPTIONS: list every assumption you made while solving 3. CONFIDENCE: rate your confidence 1-10 and explain why Task: [DESCRIBE YOUR ACTUAL TASK HERE]
What this prompt doesThe prompt asks the model to solve the task and format the answer in three sections: solution, assumptions, confidence. Replace [DESCRIBE YOUR ACTUAL TASK HERE] with a description of your real task - both models understand Russian, so you can write in Russian. - 3.Copy Carbon's response into a separate document. Note: how much time passed before the first token, how detailed the solution turned out, and whether there are errors in the code or logic.
Carbon is usually stronger in tasks involving the terminal, shell automation, and multi-step operations, but may give less precise answers on complex code reasoning.
Google AI Studio ↗ - 4.Open Anthropic Workbench and select the Claude Fable 5.2 model. Paste the exact same prompt - word for word, with the same task description.
Workbench is available after signing up in Anthropic Console. Using Fable 5.2 requires API credits - new accounts may receive initial free credits, check the pricing page.
Anthropic Console Workbench ↗ - 5.Compare the two responses by three criteria. Quality - whose solution is more accurate and complete, whether there are factual errors. Speed - who produced the first token faster and who finished the response sooner. Price - at equal quality, choose the cheaper option: until November 30 Carbon costs $2 per million input and $10 per million output tokens, Fable - $8 and $40; after November 30 Carbon rises to $4 and $20, the gap narrows.
If the task is complex code changes and reasoning, Fable is usually more accurate. If it is automation, «test - fix» loops, terminal work - Carbon is often better and cheaper with High effort until November 30.
- APIGoogle AI Studio (Carbon model)There is no free plan for Carbon. The model is available on the paid API: until November 30 - $2 per million input and $10 per million output tokens, from December 1 - $4 and $20. Also available to Google AI Ultra subscribers from $99.99 per month.
- APIAnthropic Console (Fable 5.2 model)Pay-as-you-go: $8 per million input and $40 per million output tokens. New accounts may receive free credits - check the Anthropic pricing page.
Why today
AI services change fast - interfaces and free limits may differ from what's described.