That's just a 4k context too. At a realistic context window of 16-32k tokens, the comparison becomes downright unfair.