Once again, show something that highlights both your artistic ability and your modeling ability. Don't just throw in a crate that is texture mapped with a photo of a crate. Inorganic models such as a gun or a boxy car don't show a strong modeling ability.
It's always easier to evaluate art if it includes a human with realistic anatomy. Human figures are exponentially harder to model than buildings. If you plan on showing off your modeling abilities, ask yourself how hard it would be to model that object, not how similar to your favorite game the final art looks. If you are showing off your ability to create textures, be clear that that is your goal, and be clear that you created the textures.
A common misconception is that a modeler can hide lack of skill. The best artists can model and texture human figures. There are some exceptions to the rule. I've seen some great environment artists who aren't good at modeling and texturing people, yet they have become exceptionally good at environments.
However, I would caution any artist who thought of themselves as the exception. Most artists choose to model a bush, building, or a landscape because they aren't capable of modeling a character. It's obvious when this happens.
A crazy-high polygon Zbrush model can show artistic ability but it can't be used in a game without more work. When actually making games, super high-resolution models need to be reduced and other techniques (such as normal mapping) may need to be used before they can be used in-game. It would be much more convincing to show a high resolution model along with a game-ready version that still looks great.
Animators also can deliver less than the best portfolio by including animations that aren't the type of animations often used in the game industry. Job number one is, of course, demonstrating your animation skill. However, just like other disciplines, including animations that look like they were used in games is much more convincing.
I can almost hear artists reading this right now who are arguing that many types of animations are used in games, because of cutscenes. This is a valid argument if you are applying for a company who does extensive in-house cutscenes and creates a whole new set of assets and animations that are used exclusively for those cutscenes. But I believe that this situation is the exception to the rule. Long scenes, full of dialog, that use facial morph targets, only show a limited range of skill. There is a specific skill required to create short animations that can blend with other animations and look good cycling.
I feel compelled to mention a pet peeve. Make sure walk cycles look great. Pay attention to the hips. Stiff hips make everything look stiff and unrealistic. Including walk or run cycles that have personality is very effective if they are good. However, bad hips are easy to spot.
It's okay to include art that is less related to the game industry in your portfolio. Art such as graphic design or photography can demonstrate your breadth of talent.
However, you need to be careful. Make sure that game art appears to be your focus. You won't get a concept art job when you portfolio consists of advertisements for the phone book -- even if they are amazing. You should also appear to be a game artist first, not a photographer that does game art on the side.
If I'm offering critiques and revealing my secrets, then I have to admit that my own personal portfolio hasn't been updated in years and would not fully meet these criteria -- but I'm not looking for a job. The portfolio advice I'm offering comes from years of experience in evaluating and hiring artists. I know what impresses me, and I'm confident that most art directors would agree. Evaluation is the easy part.
Don't be discouraged. Artists score good jobs all the time without a perfect portfolio. I have never actually seen a perfect portfolio, even when I've been profoundly impressed. However, it's awfully helpful to envision what the perfect portfolio could be when compiling art with hopes of landing the perfect job. As soon as you can't see how to improve your portfolio, that is when you're in real trouble.