#DeepSeekV4ProApproachingFable5

DeepSeek V4 Pro Nears Fable 5: The AI Race Is No Longer Just About Who Is the Smartest

DeepSeek V4 Pro's official update shows a major leap, especially in coding, reasoning, tool use, and agentic tasks. Several tests place its performance increasingly close to frontier models such as Claude Fable 5, although the results are not consistent across all benchmarks.
However, the most important story is not simply that “DeepSeek is nearly matching Fable 5.”
What is actually changing is the price of obtaining frontier-level AI intelligence.
The Performance Gap Is Narrowing, and the Value Competition Is Intensifying
DeepSeek V4 Pro officially exited the preview phase in mid-August with major improvements over its previous version. In independent Artificial Analysis testing cited by Reuters, the model scored 53 on the Intelligence Index, up from 40 for V4 Flash.
Meanwhile, DeepSeek's official evaluation data shows strong capabilities on complex tasks. In Max mode, V4 Pro recorded, among other results, 90.1% on GPQA Diamond, 93.5% on LiveCodeBench, and a rating of 3,206 on Codeforces. However, other benchmarks continue to show that no single model wins outright in every category.
This is the important shift in the AI race:
The difference between the best model and its competitors is shrinking, while differences in cost and access could become increasingly decisive.
No Longer Just Chatbots, but Agents That Work
DeepSeek is also strengthening its focus on AI agent capabilities—AI that not only answers questions, but can use tools, execute multi-step processes, and complete more complex tasks. Alongside the V4 Pro update, DeepSeek also introduced the open-source agentic software DeepSeek Harness.

This means the next competition may not be:
“Which AI provides the best answers?”
But rather:
“Which AI can complete real-world work most effectively?”
Conclusion
DeepSeek V4 Pro has not automatically defeated Fable 5 in every test. In fact, recent reports show that the model still lags behind in several evaluations.
But that is precisely the main point.
When a cheaper and more open model is able to approach the capabilities of frontier players, the AI market enters a new phase. Advantage is no longer determined solely by the highest benchmark scores, but by a combination of performance, cost, speed, access, and the ability to carry out real-world work.
DeepSeek V4 Pro may not have won the entire race yet. But it is helping change the rules of the race: the best AI does not always have to be the most expensive.