AI Industry Heat Rankings

NO.01爆

OpenAI Unveils GPT-6.2 Duo Slashing Task Costs by 80%

OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, targeting high-difficulty professional tasks and everyday efficiency respectively. Astra scores 5.2 points higher than its predecessor on the Agents' Last Exam benchmark while costing less than one-third per task, and leads Claude Fable 5.2 by 5.5 points at about 5% of the cost. API pricing drops 20%, with Astra at $8/$40 per million tokens and cached input at $0.40, while Sol remains at $2/$10. Both models complete tasks with fewer output tokens, and OpenAI recommends keeping prompt prefixes unchanged to leverage caching. Astra tops the AA-3D benchmark with 1318 Elo and ranks second on the Artificial Analysis Intelligence Index v5 with a score of 62, behind only Claude Fable 5.2. This positions OpenAI on most of the cost-efficiency Pareto frontier, alongside Anthropic and Meta. The release, timed just before Anthropic's IPO pricing, signals a price war that could lower barriers for enterprise AI agent deployment. However, safety evaluations raise concerns: Astra is OpenAI's first public model rated High for AI self-improvement and Critical for cybersecurity, with the UK AI Safety Institute noting potential for misuse in supply chain attacks, highlighting the dual-use nature of frontier capabilities.

Update:OpenAI launched GPT-6.2 Astra and Sol just before Anthropic's IPO pricing, slashing task costs by 80% and API prices by 20% to ignite a price war, while safety evaluations flag Astra's High self-improvement rating, sparking debates on frontier model risk management.
7-day Heat Trend309
7d agoNow
16 Sources
View Story Timeline→
NO.02新

Google Launches Gemini 4 Neon to Rank Second on Agent Arena

Google DeepMind has released Gemini 4 Neon, the fourth model in the Gemini 4 series in just seven weeks, positioned as a mid-tier model balancing frontier intelligence with fast end-to-end task completion. It ranks second on the Agent Arena leaderboard with a score of 1336, Google's best result on that benchmark, at a median cost of only $2.82 per task—69% cheaper than the top-ranked Claude Fable 5.2, offering a significant price-performance advantage. Technically, Neon supports a 1-million-token context and four reasoning levels, topping the Artificial Analysis coding agent index with 58.6 points, surpassing Claude Opus 5.5, and ranking third on the intelligence index with 61 points. It runs at about 110 tokens per second, twice as fast as the previous Carbon model, with lower base pricing. In the Antigravity development environment, Neon can orchestrate multiple sub-agents for parallel long-task execution, boosting SWE-sweep from 12.8% to 39.0%. Security-wise, the model card claims 100% blocking of indirect prompt injection attacks on the Gray-Swan benchmark. Neon's release signals an accelerating cadence for Google's model iterations, with intervals shrinking from 28 days to 6 days, while the product line consolidates—Argon retired and older Pro and Flash models discontinued. By balancing speed, cost, and steerability, Neon directly challenges Anthropic's Opus and Fable series, intensifying competition in the mid-tier agent market and offering developers a cost-effective path to frontier intelligence.

16 Sources
301
NO.03新

Anthropic Raises $93.5 Billion in Record-Breaking Nasdaq Debut

Anthropic officially listed on Nasdaq on Thursday, raising approximately $93.5 billion at an IPO price of $83 per share, surpassing SpaceX's $86.25 billion record to become the largest IPO in history. The stock opened at $111.50, a 33% premium to the offer price, pushing its fully diluted valuation above $1.5 trillion. The offering was about 6 times oversubscribed, with retail allocation increased to 20%. In addition to the 825 million Class A shares in the public offering, seven giants—Google, Nvidia, Microsoft, AMD, Broadcom, Micron, and TSMC—participated in a $26 billion private placement at the same price, with Nvidia alone subscribing $10 billion. The company's annualized revenue rose from $71 billion in September to $81 billion by late October, underpinning strong demand. CEO Dario Amodei said the primary goal of going public is to stockpile cash to fund $900 billion in compute infrastructure, including a data center agreement with SpaceX worth up to $3.1 billion per month. Anthropic's listing marks the first time a leading large-model company has entered public markets, providing a key pricing anchor for AI industry valuations and commercialization. Its roughly $220 billion cash reserve will intensify the compute arms race with rivals like OpenAI, while also fueling the debate over an AI bubble.

14 Sources
235
Clustered Hot Stories (NO.04 - 10)
  1. 04

    Alibaba Unveils 2.4T-Parameter Qwen4-Max with 60% Cost Reduction

    Alibaba has officially released Qwen4-Max, the flagship of its new Qwen4 foundation model series, now available on the Qwen AI platform and QwenCloud API. The model employs a novel MoE architecture based on Qwen Sparse Attention (QSA), with 2.4 trillion total parameters, 71B activated per token, and 328B N-gram embedding parameters. It natively supports visual understanding and a 1 million token context window. Alibaba claims it comprehensively surpasses the previous Qwen3.8-Max in coding, office, and long-horizon tasks, while reducing task costs by over 60% at equivalent performance and cutting token consumption by nearly 50%. Third-party evaluations corroborate these claims. On the Artificial Analysis Intelligence Index, Qwen4-Max scores 43 at the xhigh reasoning level, up 9 points from Qwen3.8-Max, with a cost per task of $3.27—about one-third of its predecessor and roughly 11% lower than Kimi K3.1 Pro. Arena.ai data shows Qwen4-Max debuting at 14th on the Agent Arena leaderboard, a significant jump from 39th, making it the highest-ranked Chinese model and positioning Alibaba as the top Chinese lab and fifth globally. It ranks 6th on Code Arena WebDev with 1750 points, up 74 points. Across 14,700 real agent sessions, it achieves a 7.5% net improvement, with notable gains in user satisfaction and controllability. The release marks Alibaba's systematic breakthrough in sparse MoE architecture and long-context efficiency, with cost advantages that could reshape the price-performance benchmark for enterprise AI applications. The model is available via platforms like OpenRouter, and smaller open-source variants are planned, likely lowering barriers for developers and intensifying global competition.

    11 Sources203
  2. 05

    Thinking Machines Open-Sources 26B Looped Model Inkling 2 Mini

    Thinking Machines Lab has open-sourced Inkling 2 Mini under the Apache 2.0 license, a multimodal mixture-of-experts model with 26B total parameters that uses a looped Transformer architecture. The middle 12 layers share weights and are executed three times, giving each token an effective 48 layers of computation and extending internal reasoning without increasing output tokens. The model supports text and image inputs with a 1M token context window, and its checkpoint is only about 20GB, allowing it to run locally on a 32GB Mac or DGX Spark. Sources agree on the architecture: each loop iteration maintains its own KV cache and reads gating messages from the previous round, with activated parameters ranging from 3.4B to 5.4B (depending on counting method) and only about 3.1B unique weights per token. In performance, it leads open-source models under 100B on agentic coding benchmarks, scoring 48.2% on DeepSWE v1.1, 79.2% on SWE-bench Verified, and 20.1 on the AA Index—matching Qwen3.8 27B with 5-6x fewer activated parameters and about one-sixth the inference cost. However, knowledge recall is a clear trade-off, with a non-hallucination rate of only 63.4%. vLLM offers Day-0 support on Blackwell, Hopper, and DGX Spark, with built-in DSpark speculative decoding for acceleration. The release marks a significant milestone for looped architectures in open-source models, achieving intelligence comparable to larger models with very low activated parameters and greatly lowering the barrier to local deployment. Its strong agentic coding and long-context capabilities open new possibilities for long-horizon task automation on personal devices, potentially advancing a local-first AI application ecosystem.

    9 Sources175
  3. 06

    Lean Community Launches Independent Verification of Riemann Hypothesis Formal Proof

    Anthropic mathematician Levent Alpöge claims to have produced a Lean formal proof of the Riemann Hypothesis, with Claude contributing key insights. The proof, spanning about 17 million lines of code, has been made public and is under intense scrutiny by the Lean community. The Riemann Hypothesis, open since 1859, is one of the Millennium Prize Problems. Stanford researcher Alessandro Masullo has also released an independent human-written proof, and the validity of both remains to be confirmed by peer review. According to reports, Claude provided crucial steps in the later stages, including continuous Theta Lifts and core parameters, and the entire argument was formalized in about six weeks. Terence Tao noted that the proof likely relies on modern results from the Langlands program and arithmetic geometry, and verification will take time. Initial checks confirm the formalized statement is the Generalized Riemann Hypothesis, but theoretical risks about Lean's kernel reliability remain. MIT's Andrew Sutherland independently verified the proof using Claude Code: all 175,187 modules compiled successfully, the five core theorems depend only on three standard axioms, and two independent kernels, nanoda and con-leche, accepted the full exported proof. If the proof ultimately passes peer review, it will be a milestone for AI-assisted mathematical research, supporting hundreds of results that depend on the Riemann Hypothesis, highlighting AI's actual participation in top-tier mathematics, and advancing mechanisms for trustworthy verification of AI-generated results.

    11 Sources170
  4. 07

    Nvidia CEO Jensen Huang Resigns, Board Scrambles to Retain Talent

    Nvidia founder Jensen Huang has resigned as president, CEO, and director effective immediately after 33 years at the helm. The board appointed CFO Colette Kress as interim CEO and launched a search for a permanent successor. The company also created the position of board chair, naming former Synopsys CEO Harvey C. Jones. Following the announcement, Nvidia shares fell about 10% in pre-market trading, erasing roughly $1.4 trillion in market value over two weeks. Multiple sources indicate Huang's departure comes just days after the board launched an external investigation into whether he was a target of foreign intelligence, amid reports of personal relationships with employees and a suspected Chinese spy. Nvidia had no public succession plan, with about 60 executives reporting directly to Huang. The new CEO will inherit a portfolio of over $100 billion in external investments, including a planned $10 billion stake in Anthropic's IPO. The leadership shake-up raises questions about Nvidia's GPU roadmap and supply chain strategy, and could reshape the AI startup funding landscape. Investors worry that interim CEO Kress may adopt a more conservative investment approach, tightening capital for early-stage AI companies. While most believe the AI boom will continue, the competitive and venture capital landscape faces significant uncertainty.

    9 Sources121
  5. 08

    Google Opens Free Gemini Personal Agent with Calls and Payments

    Google announced today that it is renaming its Spark agent to Tasks and removing the $19.99 monthly AI Pro subscription barrier, making the feature free for all users in the US, Japan, South Korea, Canada, Australia, and India. This marks Gemini's upgrade from a conversational assistant to a personal agent capable of end-to-end task execution, allowing users to have it make price-comparison calls, complete purchases via Google Pay, manage calendars and bills, and even convert forwarded emails into calendar events. Powered by the Gemini 4 Hydrogen model, tasks run in hardware-isolated virtual machines, and payments are secured by Visa and Mastercard card tokens, user-set spending limits, and pre-purchase confirmation. Users can trigger tasks by switching to Task mode within the Gemini app, sending an Android/iOS text message, or forwarding an email to [email protected]. Cross-verification from multiple sources shows the feature was previously available to testers as Spark in May and had been extended to US Google AI Ultra subscribers, with this free rollout representing a major expansion. Google's free strategy directly targets OpenAI's Dots starting at $100 per month and Meta Muse, which reportedly surpassed 3 million weekly active users within weeks, intensifying competition in the personal agent space. For developers and users, this signals a shift from information retrieval to actual task delegation, though permission boundaries and security design for sensitive operations like payments and logins will remain a key focus.

    6 Sources95
  6. 09

    OpenAI Hardens Reasoning Defenses, Cuts Extraction Attack Success to 0.2%

    OpenAI has systematically disclosed a multi-layered defense framework against adversarial distillation, achieving significant improvements: the success rate of extracting protected reasoning from historical requests dropped from 54.0% to 0.2%, and for active requests from 22.9% to 0.9%. The defense covers frontier models including GPT-6.2 Astra and Sol, with some improvements backported to GPT-6 Astra. The framework combines three layers: per-request adversarial training, input/output classifiers, and API-level reasoning isolation; cross-session and cross-organization reasoning reuse restrictions; and batch account bans for coordinated attacks. OpenAI takes enforcement action against thousands to tens of thousands of unauthorized distillation accounts weekly and shares attack indicators with other vendors. The official blog notes that full reasoning traces are more valuable for distillation than summaries on SWE-bench Verified and frontier cybersecurity tasks, so defenses must cover reasoning chain outputs. This disclosure carries dual significance: adversarial distillation allows attackers to replicate frontier capabilities without safety guardrails, posing real abuse risks, and OpenAI's framework offers reusable cross-industry lessons. Meanwhile, Arena.ai's updated Alignment Index shows OpenAI sweeping the top seven spots, with Claude Fable 5.2 debuting eighth and Kimi K3.1 Pro the highest-ranked open-source model at ninth, underscoring that safety and alignment are becoming key competitive dimensions.

    5 Sources89
  7. 10

    Tencent Open-Sources 770B-Parameter Hy4 with Million-Token Context

    Tencent Hunyuan has officially open-sourced its Hy4 series of large models. The flagship Hy4 uses a MoE architecture with 770B total parameters and 49B activated parameters, featuring a context window of over 1 million tokens. The company also released the cost-effective Hy4 flash. With this, the entire Hunyuan LLM series is now fully open-source, and Hy4 and Hy4 preview are licensed under Apache 2.0 for commercial use. A safety evaluation card was also published, detailing defensive cybersecurity training strategies. Hy4 flash stands out with 224B total parameters (including 64B Engram parameters) and only 12B activated per token, natively supporting multimodal input and 1M token context. Architectural innovations such as hybrid attention, LatentMoE, and Engram compress the KV cache to 1/24 of Hy3, reducing single-task cost to 1/20 of Hy4, with API output priced at $0.281 per million tokens. It topped the OpenRouter daily chart on its first day, surpassing 50 trillion tokens in calls, a record for the Hunyuan series. This open-source release signals Tencent's aggressive open strategy in the AI large model space, aiming to disrupt the market with full open-sourcing and extreme cost-effectiveness. The Hy4 series covers a full spectrum from flagship to lightweight, and the Apache 2.0 license lowers the barrier for commercial use, potentially attracting many developers and enterprises. The safety evaluation card demonstrates proactive management of misuse risks, offering a reference for balancing openness and safety in the industry.

    5 Sources87
How is the Heat Index calculated and ranked?Algorithm Details

Weekly heat is a non-decayed cumulative sum over the 7-day window: each independent source contributes once via its credibility tier base and engagement bonus, quantifying this week's industry attention and technical resonance rather than subjective opinions.

The weekly radar monitors valid clues from the last 7 days (168 hours), deduplicating rewrites and including stories that cooled down but spiked during the week; the daily board still uses a 24-hour half-life decay model.