Usual preface, Iām with PreciseBits. So while I try to only post general information take everything I say with the understanding that I have a bias.
You probably have a couple of issues here.
As stated by multiple people your surface speed is too high and that will kill you right out of the gate with steel. Generally with 304 you are going to be somewhere around 250 SFM ~3,800 RPM for a 1/4" and ~7,600 for a 1/8" (as Michael indicated). If you want more info on the why of this see here:
After that you have chipload. If this is a true metal cutter and you donāt have other issues (see below) you should be somewhere around 0.0003"-0.0005" for a 1/8" and 0.0005"-0.0010" for the 1/4". You need a big enough chip to sink the heat into. But you donāt want to either overwhelm the cutting flute or introduce deflection. You could be over double that for a steel cutter and double again for high end cutters. But Iām going to assume āgenericā for now (huge difference between āmetalā cutters and āsteelā cutters, plus geometry, carbide grade, etc.).
The other thing that can straight up kill you is runout. Runout is basically the amount you are spinning off the central axis of the spindle. In most cases spindles are pretty good. Collets are mixed though. For metal cutting you are massively fighting an uphill battle if your runout is over 0.0002". For steel you ideally want half of that. Thereās multiple reasons for this but the big one that will kill you here is that runout adds and subtracts from chipload for different flutes (itās wobbling toward and away from the cut direction while cutting).
To use these together in an example. Letās say we are using the 1/8" 4 flute cutter at the 7,600 RPM and the 0.0003" chipload. That works out to a feed of ~9 IPM. Now letās say we have 0.0002" of runout. What will happen there is that for at least part of the cut we will be cutting at a 0.0001" chipload on one flute and a 0.0005" cut on another. Or a simpler way to think about is that for a programed 9 IPM cut we will be functionally cutting at 3 IPM and 15 IPM. That can put us in a position where we are cutting both too fast and too slow.
So what I would do is measure the runout to make sure you are within margins for your cut (If not look for graded collets). Then use the higher numbers at a very low pass depth (0.005"-0.010"). This lets you extend the chipload range while limiting the force and compensating for a small chip. You can only compensate so much though. If you are successful you can tweak these for production.
As others have said you will have a much easier time cutting the with a cutting fluid. Preferably an oil based one with a surfactant. Also make sure that you are choking the tool up as much as possible without either bottoming out the tool in the spindle or inserting any ground portion of the tool in the collet.
One final thing. Those tools that were glowing are dead, even if they look okay. At the temps to make it visibly glow you are well above the temperature to leach the cobalt out of the carbide. That will greatly embrittle the carbide and that flute will not have the strength to cut steelā¦ or probably even wood.
Hope thatās useful. Let me know if thereās something I can help with.