The FX3 and the FX30 write the same files. Same XAVC S, S-I and HS codecs at the same bitrates, same two CFexpress Type A slots, same 16-bit RAW out of the HDMI port. A card from either camera goes through the same ingest and conform, and the price gap buys nothing at that stage.
The gap shows up later: in the noise reduction pass on an interview shot at ISO 6,400, in the slow-motion insert that arrived 1.6x tighter than the wide it has to cut against, in the skew on a whip pan the VFX artist has to track. Those moments are what the extra money buys.
Last updated: October 2026. Some links below are affiliate links. Buy through them and it helps keep the site running, at no extra cost to you. Prices were read on the live Amazon listings on October 12, 2026.
TLDR
Buy the FX3 at $4,298 if the work is low light, or if any of it will be tracked, stabilized or sped up in post. The second base ISO at 12,800 and the 8.8 ms readout at 4K24 are the two numbers the FX30 cannot match, and both save hours in the grade and the VFX pass. Buy the FX30 at $2,098 if the work is lit, cut at 60p or slower, and delivered as XAVC. The files are identical, the 26MP Super 35 sensor oversamples 4K from 6K, and the money saved covers the XLR handle, two 160GB cards and most of a lens.
The FX30 is our video pick in the best cameras under $2,000 for that reason. The FX3A refresh in 2025 changed the screen and the ports and nothing about the files, so the 2021 body and the 2025 body are the same decision.
Same codecs, same cards, same files
XAVC S-I at 4K24 runs 240 Mbps on both cameras, 108 GB per hour, and the recording menus match line for line from there. XAVC S is long-GOP H.264 at 100 Mbps for 4K24 (45 GB per hour), 140 at 30p and 200 at 60p. XAVC S-I is All-Intra H.264 at 240, 300 and 600 Mbps for 24, 30 and 60p, so 108, 135 and 270 GB per hour. XAVC HS is long-GOP H.265 at 200 Mbps for 4K60 and 280 for 4K120, 90 and 126 GB per hour. Every mode is 10-bit 4:2:2. Both take two CFexpress Type A or SD UHS-II cards, write proxies beside the main clip, run a fan with no record limit, and embed gyro data for Catalyst. The card side is in the CFexpress vs SD guide; S-I at 60p wants Type A, and a Lexar Gold 160GB Type A card at $210 holds about 35 minutes of it.
RAW is the same arrangement on both. Neither records it internally. Both output 16-bit RAW over HDMI to an Atomos Ninja as ProRes RAW, and since firmware v7.00 on the FX3 and v6.00 on the FX30, as Blackmagic RAW to a Video Assist. The FX3 sends 4264 x 2408; the FX30 sends 4672 x 2628 from its denser sensor, at a 1.3x crop of the Super 35 frame. The recorder decides the container, so the RAW clip that lands in Resolve looks the same from both.
For the edit this is one pipeline. S-I cuts natively in Premiere and Resolve on any machine from the last five years. HS needs hardware HEVC decode (Apple silicon, recent Nvidia and Intel parts all have it) or a proxy pass, and the in-camera proxies make that pass free. Both bodies wrap every XAVC mode in MP4, so an Avid pipeline treats them alike on ingest. The edit-ready camera guide breaks down what each codec costs in decode time; for this pair the cost is identical, and the cheaper body gives up nothing here.
Low light and dual base ISO
The FX3‘s second base ISO is 12,800. The FX30’s is 2,500. That is the largest difference between the two and it comes from the sensor: 12.1 million pixels across full frame against 26 million across Super 35 gives each FX3 pixel roughly five times the area, and Sony rates the FX3 at 15+ stops in S-Log3 against 14+ for the FX30. Between the two bases the gap is narrow; above the second base it is a different camera.
In the grade that gap is render time. The FX30 at 2,500 is clean. The FX30 at 6,400 is a noise reduction job, and temporal NR in Resolve is one of the few nodes that still pulls a 4K S-I timeline below real time on a current Mac. A dim interview or a street exterior after dusk from the FX30 is a pass the FX3 footage skips, because a 12,800 clip from the FX3 grades about like a 2,500 clip from the FX30.
The FX30 covers everything that is lit. At base 800 under a key, its 6K oversample gives a 4K frame slightly crisper than the FX3’s 4.2K readout, and the spare resolution leaves room to punch in 20 percent on an interview without going soft, a reframe the 12MP FX3 cannot offer. Lit interviews and daylight exteriors come out the same from either body, and on those jobs the FX3 is paying for a night the job never has.
4K120 and the crop
At 4K 100p and 120p the FX3 crops about 1.1x and the FX30 crops 1.6x, per Sony’s help guides. Up to 60p both read the full sensor width, the FX3 from a 4.2K readout and the FX30 oversampled from 6K. Past 60p the FX30 drops to a window roughly 4K wide in the middle of its Super 35 sensor, and a 24mm lens that framed like a 36mm at 24p frames like a 58mm at 120p. At FHD 240p the FX30 crops 1.5x, its RAW output carries the 1.3x crop noted above, and Active SteadyShot adds a further 1.13x on top of any of them.
On the timeline that is a matching problem. A 120p shot from the FX3 drops into a 24p master and cuts against the wide with no scale. The FX30’s 120p shot arrives 1.6x tighter, and the editor either accepts the jump in field of view or asks for the master to be recut around it. Interview-and-b-roll work rarely notices. Music, sport and product work, where the slow-motion shot is the shot, notices every time.
The FX30 answers this with a lens. A 16mm at 120p frames like a 38mm on full frame (16 x 1.5 x 1.6), so a shooter who plans for the crop hands post a file that needs no scaling, and the file itself is still 3840 x 2160 at 280 Mbps.
Rolling shutter
The FX3 reads a 4K24 frame in about 8.8 ms. The FX30 takes about 16 ms. For scale, the FX2 Sony shipped in 2025 reads in 27.7 ms, so the FX30 sits in the middle of Sony’s own range and the FX3 near the top of it.
Sixteen milliseconds is enough to lean verticals on a whip pan and to wobble a handheld frame during a hard footstep. On a cut-and-deliver job the eye forgives it. On a VFX job it decides whether the camera tracker solves clean or needs a rolling shutter correction before the solve. Resolve, After Effects and SynthEyes all have one, and each of them costs a render and a second look, on every shot with a screen replacement or a set extension in it.
The FX30 covers the rest of the pan. Tripod work, gimbal work at a walking pace and a handheld interview all read clean at 16 ms, and a job with no tracking in it never meets the FX30’s readout.
Audio, handle, battery
The FX3 ships with the XLR-H1 top handle in the box. The FX30 body alone does not: the handle kit adds it at $2,498, and the XLR-H1 bought on its own runs $698. The handle adds two XLR/TRS inputs with phantom power, a 3.5mm input and four channels written into the clip, which for post means a lav and a boom already in sync on the same file, with no separate recorder to line up.
Battery goes the other way. Both run on the NP-FZ100. The FX30 records about 115 minutes of 4K per battery against about 95 on the FX3 (175 against 135 by the CIPA figure), the smaller sensor drawing less. Across a ten-hour documentary day that is one fewer swap and one fewer split clip, and the FX30 is 70 grams lighter on the rig.
Price the FX30 as the handle kit whenever audio is part of the job. The kit at $2,498 still lands under the FX3 at $4,298, and the difference is a Lexar Gold 160GB Type A card plus a good part of a lens. The run-and-gun documentary kit builds the whole bag around the FX30 handle kit for that reason.
What changed in 2026
Nothing about the files. The FX3A, Sony’s May 2025 refresh of the FX3, keeps the sensor, the codecs and the price. It swaps the 1.44M-dot rear screen for a 2.36M-dot one, puts the BC-ZD1 charger in the box, drops IR and NFC, and runs the same firmware line as the original (FX3 v7.02 and FX3A v2.00 as of March 2026; the FX30 is on v6.01). Amazon’s FX3 listing is the FX3A body now, and for this comparison that is the same camera. No FX3 II or FX30 II has been announced.
The movement is above them. The FX5 was announced in July 2026 at $4,899, slipped from its August ship date to mid-October, and slots between the FX3 and FX6: a 16.6MP stacked full-frame sensor, internal 16-bit X-OCN RAW, open gate, 5K60 and 4K120. RAW on the FX3 and FX30 means an external recorder and a ProRes RAW or BRAW pipeline; RAW on the FX5 is a card, and at launch its two compressed X-OCN flavors opened only in Sony’s RAW Viewer. Our FX5 post covers what that does to a post workflow. The FX5 is the body to look at when internal RAW is the requirement; neither camera here is a step toward it.
Spec table
| Spec | Sony FX3 | Sony FX30 |
|---|---|---|
| Sensor | 12.1MP full-frame BSI | 26MP APS-C (Super 35) |
| Dual base ISO | 800 / 12,800 | 800 / 2,500 |
| Dynamic range (Sony) | 15+ stops | 14+ stops |
| 4K up to 60p | Full width, 4.2K readout | Full width, 6K oversample |
| 4K 100/120p crop | ~1.1x | 1.6x |
| Rolling shutter, 4K24 | ~8.8 ms | ~16 ms |
| Internal codecs | XAVC S / S-I / HS, 10-bit 4:2:2 | Same |
| XAVC S-I 4K24 / 30 / 60 | 240 / 300 / 600 Mbps | Same |
| RAW out (HDMI, 16-bit) | 4264 x 2408 | 4672 x 2628, 1.3x crop |
| Media | 2x CFexpress A / SD UHS-II | Same |
| Battery, 4K record | NP-FZ100, ~95 min | NP-FZ100, ~115 min |
| XLR-H1 handle | Included | Handle kit or separate |
| Weight (body) | ~715 g | ~646 g |
| Price | $4,298 | $2,098 body, $2,498 kit |
In conclusion
Two numbers separate these cameras once the footage is in the timeline: a second base ISO of 12,800 against 2,500, and a readout of 8.8 ms against 16. Everything else that reaches the edit bay, codecs, cards, proxies and the RAW path, is the same. The FX3 is for work that happens in the dark or ends up in a tracker, and its price comes back as noise reduction renders and tracking solves that never run. The FX30 is for everything lit and everything cut at 60p or slower, which is most of the work most editors receive.
Buy the FX30 handle kit unless the job description includes night, tracking or 120p, and spend the difference on glass and cards. If the FX3’s second base is the thing that sold it, confirm the shooter will use it: a lit set at ISO 800 produces identical work from either body, and the FX30 did it for less.
