Recognize that a test of general cognitive ability isn't going to test for math talent specifically, although tests that load fluid reasoning (flexible problem solving ability) tend to correlate pretty well with math achievement (tests that measure crystallized intelligence (stuff you know) and working memory (ability to hold stuff in mind and work with it on the fly) are also relevant both to being a good arithmetist (okay, I made that word up) and to being a good mathematician.

Tests that measure current math achievement also don't tell you whether that achievement is due to "talent" or "hard work" or a combination of both. Sometimes you'll see very good arithmetists who are just good arithmetists, but not very good mathematical thinkers -- they're just very good at doing what they're taught. But math achievement tests as Grinity suggests will at least tell you about what the current level of math achievement is, which is often the most relevant information for math placement.

Personally, if the question is about math talent and appropriate teaching styles for a kid, I'd use some of both -- I want to see the cognitive underpinnings that will make excellent mathematical thinking possible, as well as how the actual kid is developing in their use of those tools. I'd use both an IQ test as well as the KeyMath-3 (second choice for a math-specific referral would be the WJ-III Achievement). I like the KeyMath a lot for its comprehensive coverage of the curriculum and its focus both on acquired skills and on problem-solving techniques. It gives me a very nice picture of a kid's approach to math.

The tests I use (1:1 administration) generally have routing procedures so that a kid is not spending too much time on items at the "wrong" level of difficulty. Stuff designed for group administration is not designed that way -- usually, if the goal is GT identification, you'll do something like use the test 2 grade levels up ("out of level testing").

However, recognize that a school will tend not to do anything in much depth. They're not really trying to answer questions about learning style -- they're just trying to figure out which kids they're going to put in the seats in their existing program. Sigh.