Chat model · Google ·
Gemini 4 Carbon is a chat model by Google. In Endless it costs 307.7 credits per 1M input tokens and 1538.5 per 1M output tokens, on the Hobby plan and up.
Gemini 4 Carbon is Google's strongest reasoning model, built for hard multi-step coding, long-running agent work and cybersecurity defense.
307.7 / 1538.5 credits per 1M tokens
Gemini 4 Carbon is Google's frontier reasoning model, announced on October 28, 2026 and opened to developers and consumers on November 10. It reads text, images, PDFs and video across a 1M-token context, lets you set how much it reasons before it answers, and supports tool calling and structured outputs.
Google built it for the hardest jobs: tracking down subtle bugs across large codebases, multi-step agent work that has to stay on plan, and finding and fixing security vulnerabilities. It spends more tokens than Argon to explore, reason and check its own work. Google reports it makes about half as many unsupported claims as Argon and raises ambiguities with you instead of guessing.
Gemini 4 Argon remains the better choice for most everyday coding and writing, and Gemini 3.8 Flash the faster one. All of them live in Endless chat, and you can switch between them in the same workspace.
Start a chat in your workspace and choose Gemini 4 Carbon in the model picker.
Attach files, mention a project with @, or ask it to search the web and social media.
Generate images and video, write documents and save what matters to a project.
by Google
Very long outputs for large code migrations and deep document work
354.6 / 1773.1 credits per 1M tokens
by Google
Agentic coding and multi-step reasoning at Flash speed
133 / 664.9 credits per 1M tokens
by OpenAI
Fast visual reasoning, computer use and natural writing at low cost
88.7 / 443.3 credits per 1M tokens
Gemini 4 Carbon is on the Hobby plan and higher. Create a free workspace to try the other models first, then upgrade when you need it.