接连指控中国模型蒸馏后，Anthropic藏起Fable 5思考摘要
Related
Anthropic's trip to Washington to negotiate a lifting of the ban failed, and Claude Fable 5 will maintain its global ban.
According to Beating, the U.S. Department of Commerce and Anthropic held an emergency working meeting on Monday (June 15) regarding the recently implemented export controls on AI models. However, the meeting ended without results, and Claude Fable 5 remains globally blocked. The U.S. government's earlier restrictions stemmed primarily from security concerns, specifically the fear that if Fable 5's security defenses were bypassed through jailbreaking, it would degrade and unleash powerful cyberattacks and vulnerability exploitation capabilities comparable to Mythos, potentially leading to its misappropriation by other countries' military intelligence agencies. The closed-door meeting in Washington was led by Anthropic co-founder and Chief Computing Officer Tom Brown, Frontier Red Team leader Logan Graham, and senior security researcher Nicholas Carlini. Officials from the Office of the National Cyber Security Director and the Department of Commerce were present to listen to the discussion. The team reiterated that the U.S. government's concerns about the risks of Fable 5 jailbreaking were exaggerated, but the Department of Commerce officials and researchers from the Center for AI Standards and Innovation (CAISI) were not convinced. The government maintains that the model's security measures can be bypassed and demands that Anthropic completely resolve the jailbreak vulnerability. Currently, both sides are urgently discussing the next steps. U.S. Commerce Secretary Howard Lutnick joined the meeting remotely from the G7 summit in Evian-les-Bains, France, and maintains regular calls with Anthropic CEO Dario Amodei. Since both are attending the summit in France, Lutnick and Amodei are expected to meet face-to-face at the G7 summit this week. In response to the ban, more than 80 cybersecurity professionals, including executives from companies such as Nvidia, Adobe, and Zoom, as well as academic experts, signed an open letter on June 14th supporting Anthropic, arguing that the ban would not only harm cybersecurity defenders but also weaken the U.S. leadership in AI.
New AI assessments from Artificial Analysis show Claude is 44 times more expensive than DeepSeek.
According to Beating's monitoring, Artificial Analysis, an evaluation agency, has adjusted its AI intelligence index evaluation criteria. Instead of simply testing AI with multiple-choice questions, the new evaluation comprehensively assesses AI's ability to autonomously plan, use tools, and solve complex tasks. The new evaluation eliminates the old test of understanding simple instructions, instead introducing challenging scenarios such as simulating real-world bank customer service conversations. For the first time, the cost and time required to complete a task are included as core evaluation indicators. In the latest evaluation results, Claude Fable 5, which has been shut down by the US government, achieved the highest score of 60. Among commercially available AIs, the most expensive, Claude Opus 4.8, achieved first place with 56 points, slightly ahead of GPT-5.5 with 55 points. Domestic models also performed remarkably well, with the open-source DeepSeek V4 Pro and MiniMax M3 both achieving 44 points, followed closely by Kimi K2.6 with 43 points. The price difference between the models is significant. Running the same task using the state-of-the-art Claude Opus 4.8 costs $1.78 (approximately 13 RMB), while using the domestic open-source DeepSeek V4 Pro costs only $0.04 (approximately 0.3 RMB). This means that Claude's call cost is 44 times that of DeepSeek. The waiting time to complete a task also differs drastically; the fastest, xAI Grok 4.3, takes only 1.5 minutes, while the slowest, Claude Sonnet 4.6, takes 13.5 minutes. As the single test with the highest weight in this reform, GDPval-AA, which assesses real-world knowledge work, has been upgraded to its second version, with its weight increased to 20%. The new test sets the benchmark score for human performance at 1000 points and introduces multiple cutting-edge models to serve as judges in rotation, while also increasing the maximum number of rounds in a single dialogue to 250.
Anthropic's latest coding model, Claude Fable 5, has officially launched | _2024111120230_ | Platform
Starting Odaily 10th, developers can officially access Anthropic's newly released Claude Fable 5 model through the platform. As Anthropic's most powerful coding model to date, Claude Fable 5 is designed for challenging engineering scenarios such as legacy system migration, complex bug fixing in production environments, and long-cycle asynchronous development, demonstrating exceptional performance in code generation, logical reasoning, and problem-solving. This model further enhances its long-context understanding and task planning capabilities, easily handling complex development processes that can last for hours or even days. Currently, the API is fully open, supporting flexible integration and invocation; the web-based chat functionality will also be launched on the platform soon, providing developers with a more convenient and seamless interactive experience.
Anthropic's top models, Fable5 and Mythos5, will be released tomorrow.
According to Beating's monitoring, Anthropic issued a notice stating that the U.S. Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5, and access will resume starting tomorrow. Updates will be shared soon.
Claude Fable 5 may introduce an authentication mechanism and be billed independently of the subscription plan.
According to BlockBeats, on June 30th, AI technology analysis expert @M1Astra revealed that analysis of Anthropic Claude's application code shows the new model Fable 5 requires users to purchase credits separately for access. These credits can only be added after user authentication and are billed independently of the subscription plan. On June 27th, Anthropic announced that "the company's most powerful cybersecurity model, Mythos 5, is available for redeployment to a number of US institutions. We are also continuing to work with the government to expand access to Mythos 5 and make FABLE 5 available to the public again." --------------------------------- Click the original link below to join the Beating · Lark AI news channel for 24/7 monitoring of global AI hotspots and news.
The White House has rejected Anthropic's CEO; Fable 5 is expected to reopen after replacement negotiations.
According to Beating, US government officials have sidelined Anthropic CEO Dario Amodei in crucial negotiations regarding the re-release of the cutting-edge large-scale model Claude Fable 5. Several White House officials find Amodei difficult to communicate with, and have appointed Anthropic co-founder and Chief Computing Officer Tom Brown and Public Policy Director Sarah Heck to take over the negotiation team. This change in negotiators aims to substantially break the deadlock in advancing the security ban. The US Department of Commerce issued a security alert on June 12, pointing out a "jailbreak" vulnerability in the model and expressing concerns about foreign access to data. Anthropic subsequently removed access to Fable 5 and Mythos 5. Although Trump publicly stated in an Axios interview after meeting with Amodei face-to-face at the G7 summit that Anthropic is no longer a national security threat, substantial export controls remain in place. With Tom Brown's involvement and focus on technical communication, White House officials are satisfied with the pragmatic progress in the negotiations, significantly brightening the prospects for Fable 5's reopening.