Model launches

AI-generated text

Google makes Gemini 4 Carbon generally available, with price rise planned for December 1

Google has widened access to Gemini 4 Carbon (llm-gemini 0.36), a model that was announced two weeks ago for a limited group of cyber defenders and is now available to all users.

Google makes Gemini 4 Carbon generally available, with price rise planned for December 1

Google has made Gemini 4 Carbon (llm-gemini 0.36) generally available today. The model was announced two weeks ago and initially released to "a subset of cyber defenders"; the company has now widened access so it can be used by other developers and teams.

Pricing and scheduled increase

At launch the model is priced at $2 per million input tokens and $10 per million output tokens. Google plans to raise those rates on December 1, 2026, when prices will double to $4 per million input and $20 per million output.

Technical notes and hands-on tests

The version identifier for this release is llm-gemini 0.36. Users testing the model report that Gemini "pelican" outputs (the model’s internal labels for different thinking-effort modes) have been reliable; the default mode is medium.

In a practical test, the author used Google Antigravity IDE to prompt Gemini 4 Carbon to produce a Three.js animation. The prompt used was: "Make a three js animation where yonkers stands up as a big mech (cool stand up motion) and stomps around manhattan and queens and Brooklyn, use openstreetmap mapping data." Gemini produced a standalone HTML file (the author specifically requested to "build that into a standalone HTML file" to avoid external dependencies), and the output was captured as a video.

Update: YouTube demo of a more complex pipeline

The author also noted a YouTube video showing a more elaborate demo: using Google Antigravity’s /swarm feature, Carbon constructed an open-world 3D browser game in which the player explores Google’s Mountain View campus as a pelican riding a GBike. According to the video, the pipeline involved:

  • coordinating subagents to model 3D assets in Blender using Omni and Nano Banana references;
  • building the browser game in Three.js with WebGL shaders;
  • pairing procedural Web Audio sound effects with Lyria music.

This example suggests Gemini 4 Carbon can orchestrate multi-step, multimodal tasks that involve external tools and multiple subagents.

Why this matters

Making Gemini 4 Carbon broadly available and demonstrating integrations with tools like Antigravity IDE and /swarm shows that large language models are being applied to complex, multimodal production workflows beyond short-form text generation. The planned price increase on December 1 is also significant for developers and organizations that expect substantial token usage, as it will double the per-token cost.

Summary

  • Model: Gemini 4 Carbon (llm-gemini 0.36).
  • Availability: Announced two weeks ago for a limited group; generally available today.
  • Launch pricing: $2/million input tokens and $10/million output tokens.
  • Scheduled pricing change: $4/million input and $20/million output from December 1, 2026.
  • Demonstrations: standalone Three.js HTML animation and a YouTube-shared open-world 3D browser game prototype using subagents, Blender, Three.js and Web Audio.