- DeepSeek officially launched the V4-Flash model; Artificial Analysis named it its best-performing model in recent tests.
- Artificial Analysis estimated the cost of running V4-Flash at $0.14 per million input tokens and $0.28 per million output tokens.
- The Intelligence Index V4-Flash scored 50 out of 100, outperforming Gemini 3.6 Flash, while it showed worse results than OpenAI’s and Anthropic’s models.
China’s startup DeepSeek has officially launched the V4-Flash model. According to the Artificial Analysis company’s research, which compared the best models in tests, the difference in performance in the variant can be no more than 100 times.
Artificial Analysis suggests that V4-Flash DeepSeek costs $0.14 per million input tokens and $0.28 per million output tokens. The company also notes that the average cost of running V4-Flash in 3 tests is, for Kimi K3 from Moonshot AI — 86 tokens, for OpenAI GPT-5.6 Sol — $1.86, and for Claude Fable 5 from Anthropic — $3.15.
In explaining the rankings, Artificial Analysis points out that the “realistic” benchmark format is closer to real-world tasks, where the model’s performance depends on the number of tokens it needs to process the input and generate the output. The company also argues that the lower the token price, the better, since it is necessary to spend fewer tokens to obtain the same result.
Based on the Intelligence Index results from Artificial Analysis, which takes into account the ability to solve tasks, the model’s performance, and the cost of running it, V4-Flash scored 50 out of 100. The takeaway is that, according to the results, it outperforms Google’s Gemini 3.6 Flash and competes with OpenAI and Anthropic models; at the same time, Kimi K3 from Moonshot AI scored 57, and Claude Opus 5, Fable 5, and GPT-5.6 scored higher.
The company also notes that DeepSeek’s next version, V4-Pro, has a lower official release cost, but it is not yet known exactly how it will be priced. The company also stated that Alibaba’s Qwen3.8-Max model will have the lowest cost at the same time.