Bindu Reddy: Gemini 4 Carbon Beats Opus 5.5 for Less Money but Runs Slower
馃毃 GOOGLE ROARS BACK WITH CARBON TO OVERTAKE ANTHROPIC - Carbon outperforms Opus 5.5 - 250% cheaper than Opus - 500% cheaper than Fable - truly frontier class - not benchmark maxxed! - It totally works in the real world. The only drawback, and it's a significant one, is that it is about 3x slower than Opus even at low effort. If they manage to fix it and keep the discount, they will OVERTAKE Anthropic! Available today on ChatLLM :)
Context
A hands on comparison from a company that serves many models suggests Gemini 4 Carbon now competes with Anthropic's best on quality and cost, with speed as its weak spot.
Bindu Reddy is the CEO of Abacus.AI, whose ChatLLM product gives users access to many frontier models. Google released Gemini 4 Carbon on Nov 10 as its strongest model for complex, long horizon tasks. Claude Opus 5.5 and Claude Fable 5.2 are among Anthropic's top models.
Deep dive
Bindu Reddy posted that Google is roaring back with Gemini 4 Carbon and could overtake Anthropic. In her post she claims Carbon outperforms Claude Opus 5.5, is 250% cheaper than Opus and 500% cheaper than Fable, is truly frontier class and is not benchmark maxxed, adding that it works in the real world.
She named one significant drawback: by her account, Carbon is about 3x slower than Opus even at low effort. She wrote that if Google fixes the speed and keeps the discount, it will overtake Anthropic. She also said the model is available on ChatLLM, her company's product.
The table attached to the post lists overall scores of 84.9 for Claude Fable 5.2 at xhigh effort, 84.5 for GPT 6 Astra, 83.5 for Gemini 4 Carbon and 83.4 for Claude Opus 5.5. It shows Carbon ahead of every listed model on coding at 91.0 and on data analysis at 86.5, with a cost per successful task of $2.024, compared with $6.723 for Opus 5.5 and $9.850 for Fable 5.2. The post does not say which evaluation produced the table.
These are claims from one company's own testing and have not been independently verified.
What to expect next
Watch for Google to address speed, since Reddy calls it the main weakness, and for independent leaderboards such as LMArena and Artificial Analysis to post their own Gemini 4 Carbon results.
Price moves from Anthropic or OpenAI in response would show whether the cost gap matters to buyers. Developers will also test whether the coding lead shown in this table holds up in their own work.