Cost per finished task is the number I'd watch after Google's Gemini 4 Carbon pricing.
Cost per finished task is the number I'd watch after Google's Gemini 4 Carbon pricing. A cheaper token only wins agent and coding workloads if it finishes the job in a similar number of tokens.
- Google announced Carbon availability on Nov. 10 for paid-tier Gemini API users and enterprises, with a rollout to Google AI Ultra subscribers. Introductory rates are $2/$10 per million input/output tokens through Nov. 30.
- On Dec. 1 that doubles to $4/$20. Anthropic's Claude Fable 5.2, announced Nov. 9, lists at $8/$40. At standard uncached rates, Carbon costs a quarter of Fable now and half from December.
- Google says Carbon ranked first on several named workflow benchmarks and reports lower cost per task on those tests. Those are Google's own numbers, and independent task-cost results vary by benchmark and model settings.
- Caching, pricing tiers and reasoning-token usage all shift real spend, so a 2-4x sticker gap can shrink or grow once agents loop through a task.
If I ran high-volume agents, I'd test both on my own workload before Dec. 1, while the cheap rate makes the trial almost free. I still lean big tech wins the AI race, and $GOOG pricing like this is how it happens. Does Anthropic answer with its own price move?
Sources:
https://venturebeat.com/technology/google-gemini-4-carbon-is-now-publicly-available-and-it-undercuts-claude-fable-5-2-and-opus-5-5-on-api-price
https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/
https://www.anthropic.com/claude-fable-and-mythos-5-2
https://apidog.com/blog/gemini-4-carbon-pricing/
Source:
Apidog: https://apidog.com/blog/gemini-4-carbon-vs-claude-fable-5-2/
Not financial advice. For informational and educational purposes only.