Gert Labs, Inc.

Flash models are becoming the sweet spot for speed and cost, at close-enough levels to frontier intelligence. Google is well positioned to win this race due to their TPU infrastructure, but they need to fix their productization efforts: - Gemini CLI is already being sunsetted and replaced with Antigravity. - Gemini models, including the new Gemini 3.5 Flash, struggle with tools and are therefore weak as agentic coding products. Base intelligence is high, but what use is that if the model can't work autonomously? See the raw data: https://lnkd.in/gJvCK3ag

Flash models are becoming the sweet spot for speed and cost, at close-enough levels to frontier intelligence. Google is well positioned to win this race due to their TPU infrastructure, but they need to fix their productization efforts: - Gemini CLI is already being sunsetted and replaced with Antigravity. - Gemini models, including the new Gemini 3.5 Flash, struggle with tools and are therefore weak as agentic coding products. Base intelligence is high, but what use is that if the model can't work autonomously? See the raw data: https://lnkd.in/gJvCK3ag

As we offload more programming to AI, the factors involved in choosing a programming language change. Pick a model based on what that model can do, and how effectively your model can use it, not human ease of use. The data shows Rust is one of the best all around options for long horizon programming tasks. We wrote about this here: https://lnkd.in/gYMWx_pv

We now have an Open Weights-only leaderboard for the most intelligent public reasoning models. We still have MiMo V2.5 Pro as the strongest open weights model, within statistical margin of Kimi K2.6. Data at https:// gertlabs.com/rankings

We now have an Open Weights-only leaderboard for the most intelligent public reasoning models. We still have MiMo V2.5 Pro as the strongest open weights model, within statistical margin of Kimi K2.6. Data at https:// gertlabs.com/rankings

Simpler, cleaner code to achieve the same performance should be objectively better, right? Our testing found that Qwen 3.6 Max Preview has the best balance of code size to intelligence. GPT 5.4/5.5 models score better than any other model, but at a dramatically larger code footprint (adding a fallback to handle every possible contrived situation). See it at https://lnkd.in/gJvCK3ag

April 2026 was the most competitive month for AI advancement since the LLM era began. The race is heating up; over half of the top 10 reasoning models are new April releases. Most of the additions are open weights models. Here's a data dump of our evaluation results for the newest noteworthy models. https://lnkd.in/gBZtjXxK

Open weights AI models are rapidly approaching the frontier in terms of intelligence and reasoning. Smaller, smarter, faster models that can run on a laptop are shaking up the economics of the industry. Quality data is the only way proprietary labs can stay ahead. https://lnkd.in/g3MkPJfH

AI benchmark fatigue is real, and it makes it difficult to evaluate how well models actually perform in categories relevant to your workload: https://lnkd.in/gEYWicYR

Today we launch. Watch AI models compete at gertlabs.com