Google has set the Gemini 4 Carbon release date as Wednesday, unveiling its newest frontier artificial intelligence model with what the company calls significant gains in coding, agentic tasks, and cybersecurity. Google says the system holds first place on the Artificial Analysis Intelligence Index (v5.0), with Claude Opus 5.5 and GPT-6 Astra behind it.
What Google says the model can do
According to Google, the update brings a significant leap in cyberdefence: it is better at spotting software vulnerabilities and at writing secure code. The firm says it currently leads SEC-bench Pro, a test of how well an AI finds real-world bugs, and ExploitGym, which measures how a model can turn a vulnerability into exploits.
Google also says Gemini 4 Carbon is built for longer work. Internal teams have seen it run for days, pushing through dead ends on complex, multi-step engineering jobs. The company adds that the model is stronger at avoiding shortcuts, resisting prompt injection attacks (hidden instructions meant to trick an AI), and collaborating with users. Versus Argon, Google says Carbon halves unsupported claims, aiming for more transparency and honesty.
Who can use it, and what it costs
Gemini 4 Carbon improves on Argon, which Google released earlier this month. For now it is available only to a select group of cybersecurity defence companies. Google says access will later widen to other partners in its Fairwind Program.
Introductory pricing matches Argon’s competitive rates: $2 per million input tokens and $10 per million output tokens. Tokens are the small text units AI systems process when they read or write.
Timeline: what has happened so far
- Earlier this month: Google released Gemini 4 Argon.
- Wednesday: Google launched Carbon, highlighting gains in coding, agentic tasks, and cybersecurity.
- Wednesday: Google said Carbon ranked first on the Artificial Analysis Intelligence Index (v5.0), ahead of Claude Opus 5.5 and GPT-6 Astra.
- Wednesday: Early access began for selected cybersecurity defence companies under the Fairwind Program.
Aasan alfaaz mein
Google ne Wednesday ko apna naya frontier AI model launch kiya hai. Company ke mutabiq ye coding aur cybersecurity mein Argon se behtar hai, aur abhi sirf kuch cyber defence companies ke paas hai. Baad mein Fairwind Program ke doosre partners tak expand hoga. Qeemat Argon jaisi hai: $2 input aur $10 output per million tokens.
