Four models in fourteen days
Between 5 and 19 February 2026, four large models from three manufacturers appear — two of them on the same day. The densest stretch this review knows.
February 2026 was the month in which the pace finally became conspicuous. Not because of any single model, but because of the density.
- 5 February 2026: Claude Opus 4.6 (Anthropic) — and GPT-5.3-Codex (OpenAI) on the same day.
- 17 February 2026: Claude Sonnet 4.6 (Anthropic).
- 19 February 2026: Gemini 3.1 Pro (Google).
Two top models from competing houses on the same day — that is no coincidence but the consequence of a market in which nobody wants to watch the other take the headline alone: Anthropic: Claude Opus 4.6 (5. Februar 2026).
A confusion that comes up often
Many summaries state that «GPT-5.3» appeared in February 2026. That is only half right. On 5 February, GPT-5.3-Codex arrived, a variant geared to programming. The general GPT-5.3 only became available through the programming interface on 3 March.
Distinctions like that seem pedantic until you decide something on their basis. Anyone who selected a model in February and meant «GPT-5.3» got something other than what they thought.
What a business takes away from a month like this
The honest answer: nothing that would have to be done straight away. Four new models in two weeks does not mean your own system is four times out of date. It means a comparative test goes out of date on the day it is finished.
A model comparison from February 2026 was superseded by the end of February 2026. Not wrong — superseded.
In practice that means: the question «which model is the best?» is the wrong one. The right one is «how quickly can we switch?». Anyone who can answer that in hours never has to answer the first question conclusively.
Sources
This is how our system does it