[Maxaxle]

Before I go on, let me clarify:
1. I do NOT expect to actually find a game that fits my ridiculously strict criteria.
2. I DO expect to find SOMETHING at least mostly like what I'm looking for.
3. On the side, I'm hoping to develop a persistent GTA-style game that fits with my criteria. I have no idea what engine I'll use, but proofs of concept will most likely be done with the Blender engine. I'm not a developer, but I'm willing to learn and I've got plenty of spare time.

I am looking for a freeware, preferably open-source GTA-style game with PS1- or PS2-style graphics (full 3D in particular), half-decent AI, and at least five different cars and at least three different weapons. Multiplayer isn't a priority, but it would be nice.

Games I know of that I am NOT looking for:
-GTA1/GTA2. Technically PS1 graphics but way too basic; I'm looking for full 3D and the AI isn't great.
-Any remake of GTA1 or GTA2. Also technically PS1 graphics, but again, not full 3D, and usually no AI at all.
-TSS. No idea what the achronym stands for, but an AMAZING little probably-not-open-source, extremely basic GTA-style game. Lacks AI, for the most part, not persistent, and has zero weapons, so not what I'm looking for.
-APB Reloaded. I've been burned and burned again by this flaming turd of a game. Would be redeemable if not for the awful weapon balance, awful optimization, bland-yet-not-realistic car handling, and pay-to-win mechanics up the butt. As far as my criteria go, it's NOT freeware, it's NOT open-source, the graphics are too nice (for my purposes), and it's NOT persistent.
-Greedy Car Thieves: The devs tried to go commercial a few years back, and I'd rather not step on anyone's toes or have my own toes stepped on if the devs came back.

Engines I'm NOT looking for:
-UDK. It's resource-intensive to a painful extent. I recently tried rendering a small city (MAYBE one square mile in size, just a grid of cubes in a gigantic box), but four gigs of RAM wasn't enough to even test it in the editor. On the bright side, it's extremely user-friendly (maybe a little TOO user-friendly) and allows for persistent things everywhere.
-Quake 3. I love the possibilities demonstrated by Quake 3 Rally, Xonotic, and Smokin' Guns, but I don't have any base content to work with besides that of the aforementioned Quake 3 standalone mods and/or OpenArena. Also, I can't find anything useful as far as standalone modding, only weapon-related scripts.
-Quake 2 and 1. Same deal as Quake 3, but the physics demonstrated aren't great as well.
-GTA game engines. You would THINK that the best thing to start with would be a GTA game, but nope, there's no persistence in those. On the bright side, the modding community is so massive that it's not hard to find good mods and ways to improve the game, and piracy for GTA3 era games is so rampant that it's hard to refer to them as commercial anymore.
-GRIT. I have no idea how to use it, and right now it seems to be the most intensive engine I could get. Also, there isn't a single example of it being used by a game, for some reason.
-Panda, Unity, etc. commercial engines. Put simply, I don't want to spend a whole lot of time on an engine I don't have complete control/ownership over, and I don't want to spend money on an engine I can't work with or don't like. That particular can of worms is for someone else.
-My own engine. I have zero formal training as far as writing stuff from scratch, and my most successful work was for a Visual Basic class (which I have yet to find a good use for).