The small form-factor pluggable (SFP) is a compact, hot-pluggable optical module transceiver used for both telecommunication and data communications applications. The form factor and electrical interface are specified by a multi-source agreement (MSA) under the auspices of the Small Form Factor Committee. It is a popular industry format jointly developed and supported by many network component vendors.

An SFP interface on networking hardware is a modular (plug-and-play) slot for a variable, media-specific transceiver in order to connect a fiber optic cable or sometimes a copper cable. SFP transceivers exist supporting SONET, Gigabit Ethernet, Fibre Channel, and other communications standards. At introduction, speeds were limited to 1Gib/s, but the published SFP28 iteration is designed for speeds of 25 Gbit/s. The SFP replaced the larger GBIC in most applications, and has been referred to as a Mini-GBIC by some vendors.

A slightly larger sibling is the four-lane Quad Small Form-factor Pluggable (QSFP). The additional lanes allow for speeds 4 times their corresponding SFP. The latest published variant is QSFP28 variant allowing speeds up to 100 Gbit/s. There are inexpensive adapters allowing SFP transceivers to be placed in a QSFP port.

Both a SFP-DD, which allows for 100 Gbit/s over two lanes, as well as a QSFP-DD specifications, which allows for 400 Gbit/s over eight lanes, have been published. These use a formfactor which is backwardly compatible to their respective predecessors. An alternative competing solution, the OSFP (Octal Small Format Pluggable) transceiver is also intended for 400Gbps fiber optic links between network equipment via 8 x 50 Gbps electrical data lanes. It is slightly larger version than the QSFP formfactor which is capable of handling larger power outputs. The OSFP standard was initially announced on November 15, 2016. Its proponents say a low cost adapter will allow for QSFP module compatibility.

SFP transceivers are available with a variety of transmitter and receiver specifications, allowing users to select the appropriate transceiver for each link to provide the required optical reach over the available optical fiber type (e.g. multi-mode fiber or single-mode fiber). Transceivers are also designated by their transmission speed. SFP modules are commonly available in several different categories.

1 Gbit/s SFP

• 1 Gbit/s multi-mode fiber, LC connector, with black or beige extraction lever
• SX – 850 nm, for a maximum of 550 m at 1.25 Gbit/s (gigabit Ethernet). Other multi-mode SFP applications support even higher rates at shorter distances.
• 1.25 Gbit/s multi-mode fiber, LC connector, extraction lever colors not standardised
• SX+/MX/LSX (name dependent on manufacturer) – 1310 nm, for a distance up to 2 km. Not compatible with SX or 100BASE-FX. Based on LX but engineered to work with a multi-mode fiber using a standard multi-mode patch cable rather than a mode-conditioning cable commonly used to adapt LX to multi-mode.
• 1 to 2.5 Gbit/s single-mode fiber, LC connector, with blue extraction lever
• LX – 1310 nm, for distances up to 10 km (originally, LX just covered 5 km and LX10 for 10 km followed later)
• EX – 1310 nm, for distances up to 40 km
• ZX – 1550 nm, for distances up to 80 km (depending on fiber path loss), with green extraction lever (see GLC-ZX-SM1)
• EZX – 1550 nm, for distances up to 160 km (depending on fiber path loss)
• BX (officially BX10) – 1490 nm/1310 nm, Single Fiber Bi-Directional Gigabit SFP Transceivers, paired as BX-U and BX-D for Uplink and Downlink respectively, also for distances up to 10 km. Variations of bidirectional SFPs are also manufactured which use 1550 nm in one direction, and higher transmit power versions with link length capabilities up to 80 km.
• 1550 nm 40 km (XD), 80 km (ZX), 120 km (EX or EZX)
• SFSW – Single Fiber Single Wavelength transceivers, for bi-directional traffic on a single fiber. Coupled with CWDM, these double the traffic density of fiber links.
• CWDM and DWDM transceivers at various wavelengths achieving various maximum distances. CWDM and DWDM transceiver usually support 40 km, 80 km and 120 km link distance.
• 1 Gbit/s for copper twisted pair cabling, 8P8C (RJ-45) connector
• 1000BASE-T – these modules incorporate significant interface circuitry for Physical Coding Sublayer recoding and can only be used for gigabit Ethernet because of the specific line code. They are not compatible with (or rather: do not have equivalents for) Fiber channel or SONET. Unlike non-SFP, copper 1000BASE-T ports integrated into most routers and switches, 1000BASE-T SFPs usually cannot operate at 100BASE-TX speeds.
• 100 Mbit/s copper and optical – some vendors have shipped 100 Mbit/s limited SFPs for fiber to the home applications and drop-in replacement of legacy 100BASE-FX circuits. These are relatively uncommon and can be easily confused with 1 Gbit/s SFPs.
• Although it is not mentioned in any official specification document the maximum data rate of the original SFP standard is 5 Gbit/s. This was eventually used by the DDR Infiniband especially in its four lane QSFP form.

The enhanced small form-factor pluggable (SFP+) is an enhanced version of the SFP that supports data rates up to 16 Gbit/s. The SFP+ specification was first published on May 9, 2006, and version 4.1 published on July 6, 2009. SFP+ supports 8 Gbit/s Fibre Channel, 10 Gigabit Ethernet and Optical Transport Network standard OTU2. It is a popular industry format supported by many network component vendors. Although the SFP+ standard does not include mention of 16 Gbit/s Fibre Channel, it can be used at this speed.

SFP+ also introduces direct attach for connecting two SFP+ ports without dedicated transceivers. Direct attach cables (DAC) exist in passive (up to 7 m), active (up to 15 m), and active optical (AOC, up to 100 m) variants.

10 Gbit/s SFP+ modules are exactly the same dimensions as regular SFPs, allowing the equipment manufacturer to re-use existing physical designs for 24 and 48-port switches and modular line cards. In comparison to earlier XENPAK or XFP modules, SFP+ modules leave more circuitry to be implemented on the host board instead of inside the module. Through the use of an active electronic adapter, SFP+ modules may be used in older equipment with XENPAK ports and X2 ports.

SFP+ modules can be described as limiting or linear types; this describes the functionality of the inbuilt electronics. Limiting SFP+ modules include a signal amplifier to re-shape the (degraded) received signal whereas linear ones do not. Linear modules are mainly used with the low bandwidth standards such as 10GBASE-LRM; otherwise, limiting modules are preferred.

25 Gbit/s SFP28
SFP28 is a 25 Gbit/s interface which evolved from the 100 Gigabit Ethernet interface which is typically implemented with 4 by 25 Gbit/s data lanes. Identical in mechanical dimensions to SFP and SFP+, SFP28 implements one 28 Gbit/s lane accommodating 25 Gbit/s of data with encoding overhead.

SFP28 modules exist supporting single-or multi-mode fiber connections, active optical cable and direct attach copper.

cSFP
The compact small form-factor pluggable (cSFP) is a version of SFP with the same mechanical form factor allowing two independent bidirectional channels per port. It is used primarily to increase port density and decrease fiber usage per port.

SFP-DD
The small form-factor pluggable double density (SFP-DD) multi source agreement is a new standard for doubling port density. According to the SFD-DD MSA website: "Network equipment based on the SFP-DD will support legacy SFP modules and cables, and new double density products."

QSFP+ 40 Gb Transceiver
Quad Small Form-factor Pluggable (QSFP) transceivers are available with a variety of transmitter and receiver types, allowing users to select the appropriate transceiver for each link to provide the required optical reach over multi-mode or single-mode fiber.

4 Gbit/s QSFP
The original QSFP document specified four channels carrying Gigabit Ethernet, 4GFC (FiberChannel), or DDR InfiniBand.

40 Gbit/s QSFP+
QSFP+ is an evolution of QSFP to support four 10 Gbit/sec channels carrying 10 Gigabit Ethernet, 10GFC FiberChannel, or QDR InfiniBand. The 4 channels can also be combined into a single 40 Gigabit Ethernet link.

50 Gbit/s QSFP14
The QSFP14 standard is designed to carry FDR InfiniBand, SAS-3. or 16G Fibre Channel

100 Gbit/s QSFP28
The QSFP28 standard is designed to carry 100 Gigabit Ethernet, EDR InfiniBand, or 32G Fibre Channel. Sometimes this transceiver type is also referred to as "QSFP100" or "100G QSFP" for sake of simplicity.

200 Gbit/s QSFP56
QSFP56 is designed to carry 200 Gigabit Ethernet, HDR InfiniBand, or 64G Fibre Channel. The biggest enhancement is that QSFP56 uses PAM-4 encoding instead of NRZ. As of April 2019, this new standard has not been published, but transceivers already are in use. It uses the same physical specifications as QSFP28 (SFF-8665), with electrical specifications from SFF-8024 and the still unpublished revision 3.0 of SFF-8636. Sometimes this transceiver type is referred to as "200G QSFP" for sake of simplicity.

Ethernet switch with two empty SFP slots SFP sockets are found in Ethernet switches, routers, firewalls and network interface cards. Storage interface cards, also called HBAs or Fibre Channel storage switches, also make use of these modules, supporting different speeds such as 2Gb, 4Gb, and 8Gb. Because of their low cost, low profile, and ability to provide a connection to different types of optical fiber, SFP provides such equipment with enhanced flexibility.

The SFP transceiver is not standardized by any official standards body, but rather is specified by a multi-source agreement (MSA) among competing manufacturers. The SFP was designed after the GBIC interface, and allows greater port density (number of transceivers per cm along the edge of a mother board) than the GBIC, which is why SFP is also known as mini-GBIC. The related Small Form Factor transceiver is similar in size to the SFP, but is soldered to the host board as a through-hole device, rather than plugged into an edge-card socket.[citation needed]

However, as a practical matter, some networking equipment manufacturers engage in vendor lock-in practices whereby they deliberately break compatibility with "generic" SFPs by adding a check in the device's firmware that will enable only the vendor's own modules. Third-party SFP manufacturers have introduced SFPs with "blank" programmable EEPROMs which may be reprogrammed to match any vendor ID.