DeepSeek has rolled out an upgraded version of its V4 model, branded V4 Pro, according to Decrypt. The report compares its performance against Anthropic’s Claude Fable model.
According to Decrypt, the performance improvement is modest. V4 Pro scores only about 5% higher than Claude Fable on the benchmarks cited. That gain comes despite a price difference the outlet puts at roughly 4,500%, meaning V4 Pro costs far more to run per query or per token than its rival.
DeepSeek built its reputation as a low-cost disruptor in the large language model race. The company’s earlier releases drew attention for delivering competitive results at a fraction of the price charged by Western labs like OpenAI and Anthropic. That reputation is part of why this pricing gap, as reported by Decrypt, stands out.
If accurate, the figures suggest DeepSeek’s newest premium tier has moved away from its original cost advantage. Marginal performance gains at dramatically higher prices could complicate the company’s positioning against Anthropic, OpenAI, and other rivals competing on both capability and cost.
The AI model market has grown increasingly crowded. Labs routinely release incremental upgrades, and price-to-performance ratios have become a key metric for developers and enterprises choosing which model to integrate. A model that costs dramatically more for a small performance bump may struggle to gain adoption outside of specialized use cases where the extra accuracy matters more than cost.
For crypto-adjacent markets, AI model releases carry indirect relevance. Tokens tied to decentralized compute networks, AI-agent platforms, and machine-learning infrastructure projects often react to headline developments from major AI labs, whether Chinese or American. Traders watching AI-linked tokens sometimes treat such announcements as signals for broader sentiment toward AI infrastructure spending.
DeepSeek’s rise has also been tied to geopolitical narratives around AI competition between China and the United States. Any perceived slowdown in the company’s cost advantage could feed into ongoing debates about which country’s AI labs are best positioned for the next phase of the technology race.
As of this report, other outlets had not independently detailed the same benchmark figures or pricing comparison cited by Decrypt. Readers should treat the specific percentages as reported by that outlet, pending further confirmation from DeepSeek or Anthropic directly.
Both DeepSeek and Anthropic have not issued public statements addressing this specific comparison, based on available reporting.
Sources disagree on this story
This article was published before the reports below were compared. The reporting above stands; what follows is where the published accounts do not agree.
Decrypt and CryptoBriefing disagree on whether DeepSeek's V4 Pro update is already generally available or still a preview awaiting mid-2026 release.
What all sources agree on
- Claude Fable 5 scores about 5% better than DeepSeek V4 Pro on key benchmarks.
- Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens.
- DeepSeek V4 Pro costs roughly $0.43-$0.435 per million input tokens and $0.87 per million output tokens.
- The cost premium for Claude Fable 5 over V4 Pro is roughly 45-46 times.
Where the reports disagree
1Whether V4 Pro's update is a finished, general-availability release or still a preview
DeepSeek shipped the finished, general-availability version (0813 build) of V4 Pro on Wednesday, replacing the preview that had been running since April.
DeepSeek V4 Pro is currently available in preview with API access, and general availability is expected by mid-August 2026.
What would settle it: DeepSeek's own official release notes or API documentation confirming the status of the 0813 build.
What to make of it
Treat the roughly 5% benchmark gap and 45x cost premium figures as consistently reported across both outlets, but do not assume V4 Pro's release status is settled until DeepSeek issues an official statement or documentation confirming general availability.
Market Impact
Direct crypto market impact from this development is limited, since neither DeepSeek nor Claude Fable are blockchain-native products. However, AI-linked tokens and decentralized compute projects sometimes move on sentiment tied to major AI model releases, particularly when cost-efficiency claims are challenged.
If DeepSeek's pricing gap holds up under further scrutiny, it could shift narratives around which AI labs offer the best value, a factor that indirectly feeds into investor interest in AI infrastructure and compute-token projects within crypto markets.
The comparison highlights a broader tension in the AI sector between marginal performance gains and rising costs. Further reporting may clarify whether DeepSeek's V4 Pro pricing reflects a temporary premium tier or a lasting shift in its cost strategy.
Frequently Asked Questions
What did Decrypt report about DeepSeek's V4 Pro model?
Decrypt reported that DeepSeek's upgraded V4 Pro model performs only about 5% better than Anthropic's Claude Fable, despite costing roughly 4,500% more to run.
Why does DeepSeek's pricing matter?
DeepSeek built its reputation on offering competitive AI performance at low cost, so a large price increase for a small performance gain could affect how it is perceived against rivals.
Does this news directly affect crypto prices?
There is no direct crypto market mechanism tied to this news, but AI-linked tokens and compute projects sometimes react to sentiment around major AI model releases.
Have DeepSeek or Anthropic confirmed these figures?
Based on available reporting, neither company has issued a public statement addressing the specific performance and pricing comparison.