{"slug": "gemini-get-your-act-together", "title": "Gemini, get your act together", "summary": "Google's Gemini Carbon model, released two weeks after its announcement, delivered a time to first token of 165 seconds and throughput under 20 tokens per second in an OpenRouter request log from a coding session, according to a first-person test account published by alsado.ca. The author reported degraded performance across most of two days of testing, with one exception at 12am Mountain Time reaching roughly 30 tokens per second and a 10-second TTFT, and compared it unfavorably to Claude Fable 5.2, which the author says comes from a five-year-old company that is not a top-three cloud provider. The author concluded there is still no path to adopting Gemini models for work requiring reliable intelligence in 2026.", "body_md": "Two weeks after announcing Gemini Carbon, Google has finally made it available to users.\n\nTheir benchmarks were, as always, impressive:\n\nA model which outperforms Claude Fable 5.1 on every benchmark? I need to try this…\n\nYes, waiting two weeks after announcement to release a model in this industry is long, unheard of even. BUT, on the bright side, it’s probably a polished model now, ready to be served, **right??**\n\nThe above is an OpenRouter request log from one of my sessions, where I was trying to use Carbon for some light coding tasks and evals.\n\nIt shows the Carbon’s performance across turns, notably with a:\n\n- Time to First Token (TTFT) of 165s\n- Throughput < 20 tok/s\n\nThat’s insane. Unusable. **Embarrassing even.** I’ve never seen any model perform like this.\n\nMind you I’m using OpenRouter, which routes to the best provider at that time.\n\nThis degraded performance is unfortunately not a one-off. Across my testing in the past two days, I’ve experienced poor performance for this model during most of my testing. (One notable exception was a session I had at 12am mountain time last night where I got  *average* performance, ~30 tok/s, 10s TTFT)\n\nCompare that with Fable 5.2, a model released by a company which:\n\n- has order of magnitude less experience in infrastructure than Google\n- has been in business for 5 years (compare to Google’s 28)\n- **ISN’T A TOP 3 CLOUD PROVIDER**\n\np50 mind you…\n\nNever mind all the usual Gemini bugs, which somehow still haven’t gone away [see gemini article](https://alsado.ca/posts/gemini/). In 2026, I still don’t see a path where I could adopt Gemini models for anything requiring reliable intelligence.\n\nLooks like Bidaya’s Vertex AI credits are gonna go unused, too bad.", "url": "https://wpnews.pro/news/gemini-get-your-act-together", "canonical_source": "https://alsado.ca/posts/gemini-get-your-act-together/", "published_at": "2026-11-11 17:33:32+00:00", "updated_at": "2026-11-11 17:51:15.388326+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-infrastructure"], "entities": ["Google", "Gemini Carbon", "OpenRouter", "Claude Fable 5.2", "Claude Fable 5.1", "Vertex AI", "Bidaya", "alsado.ca"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gemini-get-your-act-together", "markdown": "https://wpnews.pro/news/gemini-get-your-act-together.md", "text": "https://wpnews.pro/news/gemini-get-your-act-together.txt", "jsonld": "https://wpnews.pro/news/gemini-get-your-act-together.jsonld"}}