For an upper stage, look to the Long March rocket series as a case history. For Long March 2-4, the boosters use hypergolic fuels while the upper stage burns LOX/LH2. Given that they have decades of experience with the hypergolic engines, that suggests LH2 is cheaper. Essentially they made the decision between building a bigger booster or adding strap-ons or going multi-core to increase payload, or else make the final stage lighter by using a higher ISP fuel. This also points out that you can't design a rocket in a (logical) vacuum. An upper stage capable of giving 3 km/s to a 10-ton payload may be cheaper to build, fuel, and operate using hypergolics, but the overall weight will be heavier than a LH2 fueled stage of the same performance, and that means boost stages must be larger and heavier and therefore cost more. A dollar spent on your upper stage may save five dollars getting it to altitude.
Also look to the J-2. Hypergolics are easy to restart. However the J-2 was restartable. Most importantly, the J-2 doesn't seem to have suffered from the same operational curses that boosters do when carrying LH2. From a purely empirical standpoint, if we flew the J-2 LOX/LH2 engines in Apollo upper stages without a lot of delays, there doesn't seem to be a compelling reason to spend more $ on a highly poisonous, heavier, corrosive replacement fuel.
That said, Long March 1 used a solid rocket upper stage. Hydrogen is tricky; development times are long and expensive.
Are you factoring system capital costs into it? R&D and building your hardware? Or is your question based on an existing sunk cost?
I am afraid that you are asking the same question that every rocket designer since Goddard has asked when they start a project with a clean sheet of paper. There is no definitive answer, only the answer which appears best, given all known factors and best modeling practices, for any specific design at a specific time.
Hypergolics are expensive and difficult to handle, but the engineering of a hypergolic engine is much simpler than for a LH2 cryogenic. So if you are looking for the fastest, cheapest system to develop and implement for a small number of launches, hypergolic is probably the better of the two. If you have more time and development money, and plan a longer service life for your system, you find yourself driven toward LOX/LH2. LOX doesn't actually appear that difficult to handle; it's the LH2 that kills you.
Per SF, $/kg of payload is your final engineering metric.
Look to history to inform your answer. Goddard and the V2 used LOX with gasoline and alcohol/water, respectively. The Titan 1 used LOX/RP1. For the Titan II, they modified the LR-87 engine into the hypergolic-fueled LR-87-5 so their ICBM could be stored with room-temperature fuel. So the decision was based on storage, not performance, and the engineering challenges were similar enough to modify a LOX/RP-1 engine rather than design something new. From this we can see that R&D and fabrication of a hypergolic engine is on par with that for LOX/RP-1 engines, which is about as cheap as liquid fueled rockets get. Hypergolic fuels are super expensive, but if your launcher has a short development cycle and limited R&D budget and you plan on a small number of launches, hypergolic wins. Actually, LOX/Kerosene wins, but that's not your question.
If you have 30 years and billions of $ to iterate your design, then LOX/LH2 wins. The proof is the Delta IV and its RS-68. If decades of engineering experience pointed to a hypergolic booster as getting payloads up more cheaply (per kg of payload to orbit), ULA would be putting money into hypergolic, or pushing the government to fund a new development effort.
I have a bias. I hate LOX/LH2 systems. LH2 is simply evil. It seeps through "cracks" in welds which any other material would consider perfectly impermeable. Hot hydrogen makes metal BLISTER. It's so cold that the foam insulation on the shuttle tanks had to be foamed with helium; foaming with air leads to the air condensing and the foam collapsing. I feel that if the shuttle program had had fewer delays due to tracking tiny hydrogen leaks, they may have been more willing to address real concerns like the SRB O-rings. I consider it an engineering miracle that they have managed to "tame" LH2 and launch the Delta IV's on schedule. Considering they are building on SSME technology, it's a miracle about 45 years in the making. Also, note that ULA only uses the Delta IV when it can't fit the payload onto a LOX/RP-1 fueled Titan, and the Delta is planned for phase-out once they get a LOX/Methane booster working.
And that's why LOX/RP-1 has regained popularity, especially in boosters. The lower ISP doesn't hurt performance nearly as badly as it does in an upper stage. Sure, it's "1950's technology", but as such it has 70 years of engineering refinement and leads to a $/kg of payload much lower than for competing LH2 systems.
Given my bias, my answer is "Neither." For a first stage, unless you have a nearly unlimited development budget and schedule, go with LOX/RP-1 or LOX/LMethane for your cheapest $/kg payload. That seems to hold true for the smallest to the largest launch systems.
Second stage? More engineering decisions, but LH2 is probably your winner. Look to the J-2 as your case history. The poor ISP of hypergolics will hurt your overall system performance more than on a first stage. Hypergolics are easy to restart. However the J-2 was restartable. Most importantly, the J-2 doesn't seem to have suffered from the same operational curses that boosters do when carrying LH2. From a purely empirical standpoint, if we flew the J-2 LOX/LH2 engines in Apollo upper stages without a lot of delays, there doesn't seem to be a compelling reason to spend more $ on a highly poisonous, heavier, corrosive replacement fuel.
And have you read up on Nitrogen Tetroxide? That stuff is evil. As far as I can tell, if you can smell it, then you are going to die.