For an A-A UCAV, you can go with either full capability, high-end ones or low-end ones.
You have to remember that UCAVs don't need a pilot or the support equipment, and that is easily several hundred kg of weight saving right off the bat. Add in a more streamlined design on account of not needing a draggy cockpit sticking out, and you would expect to get very good range for smaller birds, especially when you compare that to manned aircraft of similar size.
Now, what you are thinking off is a high-end UCAV that can go toe-to-toe with the best manned fighters and come up on top one-on-one. That is certainly one design approach. The other is to have small, cheap, expendable UCAVs.
These will carry 2-4 AAMs, probably have limited to no BVR capabilities, but make up for it with good stealth. These UCAVs will most likely work together with manned friendly fighters or AWACS, and rely on them for long range sensor support to keep weight and costs down. The friendly manned assets will using their long-range sensors to find the enemy and direct a swarm of small UCAVs in to wear the enemy down through attrition.
It is even possible that nearby manned assets might take direct control over some or all of these UCAVs. The technology is already pretty much available today. You using HMD technology to slave a camera/IRST on the UCAV to the movement of a pilots' head and feed the camera feed live back to the pilot, who would control the UCAV as if he was flying it. The close proximity should help to reduce lag to an acceptable level, and not needing to have a computer that can engage manned fighters completely automatously would drastically lower unit cost and in-service date.
The only thing that might cause issues is the bandwidth of datalinks, but provided they can achieve the needed data transfer rate, such UCAVs can be ready before the end of the decade if work began today.
I think truly automated A-A UCAVs are a couple decades away at least. And until they are ready, I do not think anyone will be making their UCAVs as expensive or large as manned fighters today.
As I mentioned earlier, UCAVs should inherently have better range than a similar sized manned aircraft, and if the weapons carriage requirements were very modest, so 2-6 AAMs, then an UCAV the size of an JF17 could have very respectable range, and may actually cost a lot less than a JF17.
I think that if Dark Sword is still an active programme and if it aims to be fielded before the end of this decade, then there is a decent chance that it will only have limited automatous combat capabilities, and that to get the most out of it, a pilot would need to take over the controls remotely. Because of this, it's size will not be much bigger than a JF17, but may have better range than a J10.