DeepSeek's V4-Flash Model is World's Cheapest AI Model
Summary
DeepSeek’s newly released V4-Flash model is being described as the world’s cheapest popular AI model to operate, with average testing costs of about 3 cents. The model trails several rivals on performance but continues DeepSeek’s low-cost AI strategy that has shaken the global tech market.
Key Points
- DeepSeek released its V4-Flash model, which researchers say is the cheapest among major AI models to operate.
- Artificial Analysis says the model costs about 14 cents per one million input tokens and 28 cents per one million output tokens, with average testing cost around 3 cents.
- The model’s performance score of 50 out of 100 matches Google’s Gemini 3.6 Flash but trails several other leading models.
- DeepSeek’s low-cost approach follows the attention drawn by its earlier R1 model, which also stirred debate over AI development costs.
