Google DeepMind Opens Gemini 4 Carbon to All: Strongest Model Targets Long-Horizon Agents and Cybersecurity

• Dispatched: Nov 11, 2026 · 00:38 ET (05:38 UTC)

One-sentence summary: Google DeepMind announced on 2026-11-10 11:00 ET / 16:00 UTC that Gemini 4 Carbon, its strongest model, is now generally available to paying developers, enterprises, scientists and consumers, targeting long-horizon agentic workflows and cybersecurity at the same introductory price as Argon.

Verification status: ✓ Official first-party disclosure (Google DeepMind official blog)

Core Facts & Consensus Delta:

Carbon was first announced on October 28, 2026, initially limited to a set of cybersecurity defenders via the Fairwind Program; the November 10 announcement is its first broad release. Benchmark performance (company-reported): ranks #1 on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench at more than 2x lower cost per task than the next best external model; ranks #2 on Artificial Analysis' new Multimodal Knowledge Agent (MKA) benchmark at more than 6x lower average cost per task than the top model; Artificial Analysis Intelligence Index score of 59. On safety, Carbon records a 0% attack success rate on the Gray Swan indirect prompt injection benchmark and a 92% pass rate on a 375-task agentic honesty benchmark, versus 70% for Argon.

Pricing is $2 per million input tokens and $10 per million output tokens during the introductory period, identical to Argon, rising to $4 / $20 after November 30, 2026. Access runs through the Gemini API paid tier (AI Studio, Android Studio) and Gemini for business; Google AI Ultra subscribers get Carbon in Antigravity, the Gemini app and AI Mode in Search; scientists can apply for six months of free top-tier Antigravity access. Google is also previewing /swarm in Antigravity, turning a single prompt into a coordinated team of agents.

Supply-Chain & Market Read-Through: Carbon lands one day after Anthropic's Claude Fable 5.2 launch (November 9, 2026, with a 20% API price cut) and on the same day Anthropic's IPO roadshow began, sharpening the contest among the leading labs for long-horizon agentic and enterprise coding workloads; Google is competing on capability plus lower cost per task. Longer verification loops and multi-agent orchestration consume more tokens per task, adding demand for Google Cloud and its in-house TPU compute; Carbon's cybersecurity capabilities — finding, exploiting and patching vulnerabilities end to end — serve enterprise security teams and the open-source ecosystem directly.

There is no immediate direct impact on memory fundamentals: the release does not change the supply-demand picture for HBM and memory makers (SK hynix 000660.KS, Samsung Electronics 005930.KS, Micron MU, SanDisk SNDK); sustained agent-workload growth would, however, support enterprise storage and HBM demand indirectly through hyperscaler compute procurement.

Next Catalysts & Watchpoints:

• December 1, 2026: introductory pricing ends; input/output prices double to $4 / $20 per million tokens — watch developer retention and usage.
• November 18, 2026 (expected): Anthropic IPO pricing; November 19, 2026 (expected): Nasdaq listing — a read on market appetite for AI-lab commercialization.
• November 17, 2026, after market close: Nvidia earnings — watch data-center guidance for signs of agent-workload expansion.

Sources:

• Google DeepMind official blog (Published: 2026-11-10 11:00 ET / 16:00 UTC)
• ExplainX: Gemini 4 Carbon launch details and Artificial Analysis Intelligence Index

Information only, not investment advice.