OM1–OM5 Grades Compared: Multimode Fiber Bandwidth, Reach, and Bend Radius

OM1 through OM5 are grades of multimode fiber, and they differ in one thing that matters: how much modal bandwidth the glass delivers at 850 nm, which is what sets how far you can run a given link speed. OM1 (62.5/125 µm) is legacy. OM2, OM3, OM4 and OM5 are all 50/125 µm — physically the same glass geometry, graded by bandwidth. The jump that actually changed data-center design was OM3 → OM4 (10G reach 300 m → 550 m). OM5 is the one most often misunderstood: it does not beat OM4 at 850 nm.

Key takeaways

• OM1 is the odd one out. 62.5 µm core; OM2–OM5 are all 50 µm. Never mix the two in a link.
• OM3 → OM4 is the real upgrade: effective modal bandwidth ≥2000 → ≥4700 MHz·km, and 10G reach 300 m → 550 m.
• OM5 does not raise 850 nm bandwidth. Measured at 850 nm it matches OM4 (≥4700 MHz·km). Its advantage is multiple wavelengths, not more speed on one.
• Attenuation barely moves across the grades — 3.0/1.0 dB/km typical at 850/1300 nm on all of them. Bandwidth is the variable, loss is not.
• Bend-insensitive multimode is a separate axis: BI-OM3/BI-OM4 drop the minimum bend radius from 30 mm to 7.5 mm at the same bandwidth grade.

OM1–OM5 at a glance

Bandwidth and reach figures below are as published on DYS fiber tables. Where a cell reads "not specified", the grade genuinely carries no such figure — OM1 and OM2 predate laser-launch specification, so they have no EMB rating at all:

OM1 and OM2 do carry short 10G reaches under IEEE 802.3 — a few tens of metres — but they are not laser-optimized and no serious 10G design starts there, which is why the grades above exist at all.

Read the last two rows again: OM5 carries the same ≥4700 MHz·km EMB as OM4. Since 850 nm reach follows from that bandwidth, an OM5 link behaves like an OM4 link on a single 850 nm wavelength. That is not an error in the table, and it is the single most useful fact on this page — see the OM5 section below.

Note the two bandwidth columns. OFL (overfilled launch) is measured with an LED-style launch that fills every mode; EMB (effective modal bandwidth) models a VCSEL laser launch, which is what actual 10G/40G/100G transceivers use. OM3 and up are specified with EMB because they are laser-optimized — that is what "laser-optimized multimode" means, and it is why OM1/OM2 carry no EMB figure.

Attenuation: the spec that does not change

A common assumption is that higher OM grades are "lower loss" fiber. They are not:

(OM5 shares the 50/125 graded-glass envelope; DYS publishes its OM5 typical figure at 3.0/1.0 dB/km.)

Identical across the board. Multimode reach is not limited by how much light the glass absorbs — over 300 m, 3 dB/km costs you under 1 dB. It is limited by modal dispersion: many light paths (modes) travel the core at slightly different effective speeds, so a pulse launched sharp arrives smeared, and past some distance the smearing closes the receiver's eye. Grading the refractive index profile more precisely is what buys OM3 and OM4 their bandwidth. So when you upgrade OM3 → OM4 you are buying less dispersion, not less loss, and your loss budget stays roughly where it was.

OM5: what it actually buys you

OM5 (wideband multimode, "WBMMF") is specified at ≥4700 MHz·km EMB at 850 nm — the same as OM4. On a single-wavelength 850 nm link, OM5 and OM4 do the same job at the same reach. Anyone selling OM5 as "faster than OM4" is selling the wrong thing.

What OM5 adds is characterized bandwidth across a range of wavelengths above 850 nm, so that a transceiver can run several wavelengths down one fiber at once (short-wavelength division multiplexing). The value is fiber count: four wavelengths on one pair does the work of four pairs. That is a structured-cabling economics argument for high-density data centers, not a speed argument.

The practical consequence for buyers: OM5 pays off only if you are actually deploying SWDM transceivers. If you are not, OM4 delivers identical 850 nm performance, and the money is better spent elsewhere. DYS multimode assemblies are available in OM3, OM4 and OM5 — the choice is a transceiver-roadmap decision, not a cable-quality one.

Bend-insensitive multimode (BI-OM3 / BI-OM4)

Bandwidth grade and bend performance are independent axes, and most comparison articles collapse them. Bend-insensitive multimode carries the same OM3 or OM4 bandwidth grade, but a trench in the refractive index profile keeps light in the core around tight turns. The difference on DYS published tables:

A 4× reduction in bend radius, at no bandwidth cost. That matters exactly where multimode lives: patch panels, cassette fanouts, high-density racks — the places where a technician's cable management is what actually breaks a link. "BI-OM4" is not a grade above OM4; it is OM4 that tolerates the rack.

OM1 and OM2: why you still meet them

OM1 is 62.5/125 µm — a wider core than everything after it. It is the reason mixing grades is dangerous. Launching from a 62.5 µm core into a 50 µm core spills the light that does not fit: roughly 3–5 dB of one-way loss at that junction, and the loss is directional, so the link can appear to work in one direction and fail in the other. Diagnosing that from a link budget alone wastes an afternoon.

• Never mix 62.5/125 and 50/125 in one channel. If you inherit a mixed plant, document the transition points.
• OM1's orange jacket looks like OM2's orange jacket. Both are orange under TIA-598. Colour will not save you here — read the print legend. See the fiber optic color codes chart for why the jacket is a weak identifier.
• OM2 is 50/125, so it is dimensionally compatible with OM3/OM4/OM5 — the penalty for mixing is bandwidth, not a core mismatch.

Neither grade is specified for modern laser links, and neither carries an EMB figure. Treat both as brownfield-only: fine to splice and maintain, wrong to design with.

Which grade to specify

• New multimode build, no SWDM plans — OM4. It is the reach/price knee, and 550 m at 10G covers most campus and data-center distances.
• Short, dense, cost-sensitive runs — OM3 if 300 m at 10G is comfortably enough. The OM3→OM4 delta is bandwidth you will not use.
• SWDM transceivers on the roadmap — OM5. Otherwise it is OM4 at a premium.
• Anything past ~550 m, or a link you do not want to revisit — stop specifying multimode. Single-mode (ITU-T G.652) has no modal dispersion limit; the cost has moved into the optics, not the fiber.
• High-density patching — add BI to whichever grade you chose. Same bandwidth, 7.5 mm bend radius.

For how core and cladding sizes underpin all of this, see the fiber optic cable size chart. For cable construction and jacket choice, start at the cable selection guides.

Frequently asked questions

What is the difference between OM1, OM2, OM3, OM4 and OM5?

Core size and modal bandwidth. OM1 is 62.5/125 µm; OM2 through OM5 are all 50/125 µm. Bandwidth at 850 nm rises OM1 (≥200 MHz·km OFL) → OM2 (≥500) → OM3 (≥1500 OFL, ≥2000 EMB) → OM4 (≥3500 OFL, ≥4700 EMB). OM5 matches OM4 at 850 nm and adds characterized bandwidth at longer wavelengths for SWDM.

Is OM5 faster than OM4?

No. OM5 and OM4 are both specified at ≥4700 MHz·km effective modal bandwidth at 850 nm, and both reach 550 m at 10GBASE-SR. OM5 adds usable bandwidth at additional wavelengths so one fiber can carry several channels at once (SWDM). It saves fiber count, not time.

Can I mix OM3 and OM4 in the same link?

Yes — both are 50/125 µm, so there is no core mismatch. The link simply performs to the weaker grade over the mixed section, so budget the reach as if the whole run were OM3. Mixing OM1 (62.5 µm) with any 50 µm grade is the combination to avoid.

What happens if you connect 62.5/125 to 50/125?

Launching from the 62.5 µm core into the 50 µm core loses the light that does not fit — on the order of 3–5 dB one way. The loss is directional: 50 → 62.5 costs little, 62.5 → 50 costs a lot, so the link can test acceptably one way and fail the other.

What is the difference between OFL and EMB bandwidth?

OFL (overfilled launch) is measured with a launch that fills all modes, representing LED sources. EMB (effective modal bandwidth) models a VCSEL laser launch, which is what real 10G+ transceivers use. OM3 and above are laser-optimized and specified by EMB; OM1 and OM2 predate that and carry OFL figures only.

Does a higher OM grade mean lower loss?

No. Attenuation is 3.0/1.0 dB/km typical at 850/1300 nm across OM1 to OM5 alike. Higher grades buy less modal dispersion, not less loss — your link budget does not improve when you move up a grade.

Is BI-OM4 better than OM4?

It is the same bandwidth grade with better bend tolerance — minimum bend radius 7.5 mm against 30 mm — not a faster fiber. It is worth specifying in patch panels, cassettes and dense racks, where tight bends are a real failure mode, and irrelevant in a straight duct run.

Is multimode still worth deploying?

Inside a building or data center, yes — multimode optics remain cheaper than single-mode at short reach, which is the whole basis of the trade. The decision hinges on distance: past roughly 550 m at 10G you are outside OM4's reach and single-mode is the answer regardless of optics cost.