Gemini and Musk keep up with OpenAI's pace, while Llama 3 and Z fall behind the standard

Radar picked up a strong signal from X/Twitter by Ethan Mollick.

Competitors outside the top tier struggle to catch up

Last week in the LLM market showed a clear trend. Google released a high-quality flash model that brings an efficient solution. Mark Zuckerberg made a similar move by announcing the new "Z" model. Additionally, Meta showed it is still moving forward with the release of Muse Spark 1.3, while GLM 5.3 is only a few weeks old.

These models confirm that development is not resting and companies are investing in continuously improving existing architectures. This creates a broader ecosystem of alternatives to the most famous tools.

The gap between the best and the rest deepens

Despite all the new models from Google or Meta, a harsh reality is emerging. Analysts and people testing the latest technologies confirm that at the moment, nothing comes close to the most powerful models like Astra or Fable in terms of capabilities.

The frontier model market (those at the very edge of technological possibilities) is currently shaping up as a race between two main players. For developers who need the highest available reasoning and context capacity, the choice is narrowing, despite the wide range of cheaper and smaller models from other providers.

Focus on efficiency instead of absolute performance is growing

While the absolute top tier pulls away, other companies are evidently focusing on the fast model segment. These are cheaper, faster, and smaller solutions that can handle daily routines. This shift means that companies outside the narrow top tier may not win benchmarks in complex reasoning, but they might win the enterprise integration market due to speed and price.

Practical adoption will reveal the winner of this split

The question remains how long this "two-company race" state will last. Attention must be directed to how much the smaller, more efficient models from Meta or Google will suffice for everyday enterprise scenarios, and how much top-tier models like Astra will maintain their exclusivity without being priced out of common use by smaller models.

Lilith's verdict

This is the receipt for the illusion of broad competition: the charts are filling up with new names, but when you need to solve a complex problem, you still call the same two numbers.