{"slug": "chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price", "title": "China’s DeepSeek Upgrades V4 Pro: Claude Fable Is Only 5% Better at 4,500% the Price", "summary": "DeepSeek quietly released the finished version of its flagship model, DeepSeek-V4-Pro-0813, on Wednesday without an announcement, replacing the preview build that had been available since April. The company's own benchmarks show Anthropic's Claude Fable 5 leads by an average of 5.3% across nine agent benchmarks, but Fable 5 costs $10 per million input tokens and $50 per million output versus V4 Pro's $0.435 and $0.87, making the premium roughly 46 times more expensive. DeepSeek's pricing remains about 98% below GPT-5 Pro, and independent labs have not yet benchmarked the 0813 build.", "body_md": "#### In brief\n\n- DeepSeek quietly swapped deepseek-v4-pro to the 0813 build, the general-availability release of a model that had been running as a preview since April.\n- Across nine agent benchmarks where both models are scored, DeepSeek’s table puts Claude Fable 5 ahead by an average of 5.3%. On two of them, DeepSeek wins.\n- Fable 5 costs $10 per million input tokens and $50 per million output. V4 Pro costs $0.435 and $0.87.\n\nDeepSeek shipped the finished version of its flagship on Wednesday with no blog post and no announcement. The tell was a table cell: the model version listed for “deepseek-v4-pro” on the [API pricing page](https://api-docs.deepseek.com/quick_start/pricing/) now reads DeepSeek-V4-Pro-0813.\n\nThe cost to use this model is around $0.435 and $0.87 per million tokens (the basic unit of information a model can handle) of input and output. So the pricing structure remains the same, but what changed is the weights underneath.\n\nDeepseek V4 Pro has been in the wild since April, priced [98% below GPT-5 Pro](https://decrypt.co/365455/deepseek-v4-launch-pro-version-costs-less-gpt-5-pro), and every independent lab that tested it was testing a preview. DeepSeek said so itself on July 31, when it pushed V4-Flash to general availability and noted that the Pro API was “unchanged” with the official release to “follow soon.” The model card on Hugging Face, the go-to repository for open-source AI projects, still describes the V4 series as “a preview version.”\n\nSo the widely circulated scores describe a build DeepSeek considered unfinished. Nobody outside the company has independently benchmarked 0813 yet.\n\n**What DeepSeek’s table claims**\n\nThe company published a comparison across 10 agent benchmarks. On the eight where Fable 5 or another model has the advantage, the gaps are small.\n\nAverage Fable 5’s relative lead across the benchmarks and you get 5.3%. Strip out Humanity’s Last Exam without tools—where DeepSeek scores 42.7 against 53.3, a 10.6% gap that skews everything—and the remaining rows average 2.8%.\n\nPricing is public on both sides, and this is where the comparison stops being close. Fable 5 runs $10 per million input tokens and $50 per million output. V4 Pro runs $0.435 and $0.87, with cached input at $0.003625. On blended rates that’s $30 against $0.65—roughly 46 times, or 4,600% of the cost. That’s the kind of spread that matters when you’re running a business and using AI tools at scale.\n\nCost per completed task runs wider, because Fable 5 thinks longer and writes more. Artificial Analysis measured it at $3.15 per benchmark task against 3 cents for V4-Flash, [about 105 times cheaper](https://www.reuters.com/business/retail-consumer/deepseeks-new-ai-model-is-by-far-cheapest-well-known-models-run-research-firm-2026-08-03/). Hugging Face CEO Clément Delangue [put the spread at over $31 per task](https://x.com/ClementDelangue/status/2067768150612193390) against roughly $0.04. No per-task figure exists for 0813 yet.\n\nAnthropic’s own lineup complicates the premium. Claude Opus 5 [outscores Fable 5 on most benchmarks at half the price](https://decrypt.co/374305/claude-opus-5-outscores-fable-5-most-benchmarks-half-price).\n\nDeepSeek scored these itself, on infrastructure it hasn’t released. Its July note specified DeepSeek Harness minimal mode “to be released soon,” running at max effort with high creativity. Two of the ten benchmarks, DSBench-FullStack and DSBench-Hard, are internal test sets with no public leaderboard to check them against.\n\nThe direction of travel is consistent, though. Chinese open-weights labs keep landing within a few points of the American frontier at a fraction of the price—[Kimi K3 beat Fable 5 and GPT-5.6 Sol](https://decrypt.co/373716/china-kimi-k3-largest-open-source-ai-model-ever-beats-claude-fable-gpt-5-6-sol) on release, and DeepSeek and Xiaomi have been [cutting frontier costs by 99%](https://decrypt.co/369202/deepseek-xiaomi-frontier-ai-cheaper-american-labs-other-way) while U.S. labs go the other way. DeepSeek’s weights are MIT-licensed and on Hugging Face, so independent verification is a download away.", "url": "https://wpnews.pro/news/chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price", "canonical_source": "https://decrypt.co/375507/china-deepseek-upgrades-v4-pro-claude-fable", "published_at": "2026-08-12 20:01:05+00:00", "updated_at": "2026-08-12 20:18:31.871680+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products"], "entities": ["DeepSeek", "DeepSeek-V4-Pro-0813", "Anthropic", "Claude Fable 5", "Hugging Face", "Clément Delangue", "Artificial Analysis", "Kimi K3"], "alternates": {"html": "https://wpnews.pro/news/chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price", "markdown": "https://wpnews.pro/news/chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price.md", "text": "https://wpnews.pro/news/chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price.txt", "jsonld": "https://wpnews.pro/news/chinas-deepseek-upgrades-v4-pro-claude-fable-is-only-5-better-at-4500-the-price.jsonld"}}