Today for AI
HOT RADAR
llmHEAT 195.2°

AI CURATED HOT STORY · 10/29/2026

Google Unveils Gemini 4 Carbon, Leading in Security and Long-Horizon Tasks

21 reports archived16 independent sourcesupdated 10/29/2026, 20:43:13
Synthesis & Latest Updates
16 Sources Cross-Validated

Google has officially launched the next-generation frontier model, Gemini 4 Carbon, establishing a new state-of-the-art (SOTA) benchmark in cybersecurity defense and long-horizon agentic tasks. The model secured the top position on the Artificial Analysis Intelligence Index v5.0 with a score of 59, achieving an exceptional 88.2% vulnerability discovery rate on SEC-bench Pro and doubling the probability of generating bug-free code compared to its predecessor.

Cross-verified data indicates that Gemini 4 Carbon supports a 1M token context window and demonstrates multi-day autonomous engineering capabilities, scoring 96.2% F1 on GraphWalks benchmarks. While slightly trailing Claude Opus 5.5 in coding agent metrics, it ranks first on the Arena Text leaderboard and offers a median task cost approximately 72% lower than competitors, redefining the performance-cost Pareto frontier.

Currently deployed via a phased access strategy prioritizing security defenders and trusted partners, the model's advanced root cause analysis and steerability signal a new era for AI in complex software engineering and security automation, significantly impacting developer ecosystems and enterprise application landscapes.

LATEST/Gemini 4 Carbon is currently in limited testing via the Fairwind program for security defenders. Priced at $4/$20 per million tokens with significant caching discounts, it receives high praise from the Chrome team for patch quality. Public access is expected to expand gradually.

Hourly heat curve

20 fully observed hours
Now
195.2
Peak
219.905:00
24h change
+18%

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 21 reports · Latest first
  1. InfoQ 中文T2·88 pts

    Google Launches Gemini 4 Carbon Topping Intelligence Index with Enhanced Security

    • Gemini 4 Carbon tops the Artificial Analysis v5.0 Intelligence Index with 59 points, surpassing Claude Opus 5.5 and GPT-6 Astra.
    • Designed for sustained agentic workflows, it excels in debugging large codebases and multi-step engineering tasks.
    • Cybersecurity capabilities are significantly enhanced with an 88.2% vulnerability discovery rate, though initial access is limited to trusted defense experts.
  2. Hacker News AIT2·85 pts

    Gemini 4 Carbon Tops Intelligence Index with Enhanced Reasoning and Coding

    • Gemini 4 Carbon ranks #1 on the Artificial Analysis Intelligence Index v5 with a score of 59
    • Leads in 4 out of 9 sub-evaluations, including HLE-Diamond, showcasing strong reasoning and scientific knowledge
    • Significant gains in coding capabilities, ranking second only to Claude Opus 5.5 but not leading all areas
    • Currently in limited rollout; standard pricing is $2/$6 per 1M input/output tokens, with a 95% discount on cached inputs
  4. 量子位官网T2·92 pts

    Google Launches Gemini 4 Carbon, Tops Intelligence Index with Price Cuts

    • Gemini 4 Carbon tops the Artificial Analysis Intelligence Index with a score of 59, surpassing Claude Opus 5.5 and GPT-6 Astra.
    • The model exhibits 'autonomous engineer' capabilities, capable of running for days to locate and fix complex underlying bugs.
    • Claims 100% immunity to prompt injection, significantly enhancing code security and anti-cheating measures.
    • Launch pricing is half that of competitors: $2/M input, $10/M output, with cached input as low as $0.2/M.
  6. IT之家 · 智能时代T2·85 pts

    Google Launches Gemini 4 Carbon for Cybersecurity and Long-Horizon Tasks

    • Gemini 4 Carbon ranks #1 globally with a score of 59 on the Artificial Analysis Intelligence Index v5.0.
    • Key strength in cybersecurity: achieves 88.2% on SEC-bench Pro vulnerability discovery and resists indirect prompt injection attacks.
    • Designed for long-horizon agentic tasks; improved WebDev Arena Elo by 89 points and reduced over-execution rate by 26%.
    • Launch pricing offers discounts ($2/M input tokens), reverting to standard high prices later; currently limited to security defenders.
  9. The DecoderT2·92 pts

    Google DeepMind Releases Gemini 4 Carbon Topping Intelligence Index

    • Gemini 4 Carbon scores 59 on the Artificial Analysis Intelligence Index v5, a significant jump from Argon's 50.
    • Positioned for multi-day autonomous development workflows, its coding agent index is close to but still trails Claude Opus 5.5.
    • Due to advanced cybersecurity capabilities, initial access is restricted to security experts undergoing U.S. government pre-testing, with pricing unchanged from Argon.
  11. Ars Technica AIT2·78 pts

    Google Unveils Gemini 4 Carbon with Major Security Gains, Still Unavailable

    • Gemini 4 Carbon is an upgrade to Argon, not a replacement for the Flash series, following a staged release strategy
    • The model excels in cybersecurity, doubling the probability of generating vulnerability-free code compared to Argon
    • Chrome team claims it generates high-quality patches 10x better than non-Google models
    • Score increased by 89 points on WebDev Arena, ranking fourth on the leaderboard
  12. 宝玉 (@dotey)T2·85 pts

    Google Releases Gemini 4 Carbon Topping Benchmarks Restricted to Cyber Defenders

    • Gemini 4 Carbon ranks first in third-party comprehensive evaluations, surpassing Claude Opus 5.5 and GPT-6 Astra
    • The model features multi-day continuous operation, self-correction, and high success rates in generating secure code
    • A tiered rollout strategy restricts initial access to national critical infrastructure and security agencies via the Fairwind program
  13. Marktechpost (@Marktechpost)T2·85 pts

    DeepMind Unveils Gemini 4 Carbon for Long-Horizon Agents and Vulnerability Patching

    • Core capability focuses on 'long-horizon agents', capable of running for days to root-cause bugs in large codebases.
    • Establishes a complete cyber defense loop: find flaws -> prove exploitability -> fix root causes.
    • Strong benchmark performance: Scores 59 on Artificial Analysis Index (surpassing Claude Opus 5.5 and GPT-6 Astra) and 54.8% on ExploitGym.
    • Restricted access and undisclosed specs: Currently available only via the Fairwind Program to trusted partners; no public API date; pricing at $2/$10 per 1M input/output tokens.
  14. Yuchen Jin (@Yuchenj_UW)T2·45 pts

    Gemini 4 Carbon Outperforms Opus 5.5 in Artificial Analysis Benchmarks

    • Gemini 4 Carbon demonstrates higher intelligence and lower per-task cost than Opus 5.5 according to Artificial Analysis benchmarks.
    • Results rely on the Antigravity CLI environment, potentially introducing bias due to inference framework differences.
    • The industry is cross-verifying actual performance using alternative tools like Codex and Omnigent.
  15. Rohan Paul (@rohanpaul_ai)T2·92 pts

    Google Launches Gemini 4 Carbon, Surpassing GPT-6 in Security and Long Context

    • Gemini 4 Carbon scores 88.2% on SEC-bench Pro for vulnerability discovery, significantly outperforming predecessors and competitors.
    • Supports up to 1M token context windows, achieving a 96.2% F1 score on the GraphWalks benchmark.
    • Demonstrates 'multi-day' autonomous engineering capabilities, performing deep root-cause analysis across large codebases without stalling.
    • Implements a phased access strategy: starting with cyber defenders and trusted partners before expanding to general users.
  16. Arena.ai (@arena)T1·85 pts

    Gemini 4 Carbon Debuts at #1 in Text Arena with Major Agent Gains

    • Gemini 4 Carbon ranks #1 in Text Arena and #2 in Agent Arena, surpassing GPT-6 Astra and Fable-5.1
    • Achieved a +16.2% net improvement in agent tasks based on over 21.9K real-world sessions
    • Median cost per task is $1.80, which is 72% lower than the #1 ranked Claude Opus 5.5
    • Stands out in key signals such as Steerability (+18.6%) and Confirmed Success (+18.7%)
  17. Artificial Analysis (@ArtificialAnlys)T1·85 pts

    Gemini 4 Carbon Tops Intelligence Index, Trails Only Claude in Coding

    • Gemini 4 Carbon tops the Intelligence Index v5 with a score of 59, ranking #1 in four sub-evaluations including HLE-Diamond
    • Strong coding agent capabilities, trailing only Claude Opus 5.5 overall, though not leading in pure code generation
    • Currently in limited rollout; standard pricing is $4/$20 per 1M tokens, with a 95% discount on cached inputs
  19. Hacker News AIT2·88 pts

    Google Launches Gemini 4 Carbon with Focus on Cybersecurity and Long-Horizon Tasks

    • Gemini 4 Carbon scores 59 on the Artificial Analysis Intelligence Index v5.0, ranking #1.
    • Excels in cybersecurity benchmarks: 88.2% on SEC-bench Pro (vulnerability finding) and 54.8% on ExploitGym (exploitation).
    • Emphasizes long-horizon capabilities, suitable for complex agentic tasks and secure code generation.
    • Clear API pricing established: $4 per 1M input tokens and $20 per 1M output tokens.
  21. Artificial AnalysisT1·88 pts

    Gemini 4 Carbon Tops Intelligence Index v5 with Major Performance Gains

    • Gemini 4 Carbon ranks #1 on Intelligence Index v5, leading in four specific evaluations including HLE-Diamond and Terminal-Bench-Science
    • Shows substantial gains in coding and knowledge work compared to other Gemini models, though still second to Claude Opus 5.5 in coding agents
    • Supports 1M token context window and multimodal input, priced at $4/$20 per 1M tokens with a 95% discount for cached inputs