H.264 vs H.265 (HEVC): Quality, Bitrate and Device Support

Evaluated October 2026: in the h264 vs h265 decision, HEVC (H.265) matches H.264 (AVC) quality at roughly 30 to 50 percent lower bitrate, with the largest savings at 4K and HDR. H.264 is still the only codec that every browser, phone, set-top box and smart TV decodes. Stream HEVC for 1080p and above, HDR and living-room devices. Keep an H.264 ladder as the universal fallback, and drop it only when your player telemetry shows the HEVC-capable share is close to 100 percent.

As of 2026, H.265 (HEVC) encoders deliver H.264-equivalent quality at about 30 to 50 percent less bitrate, with gains near 50 percent at 2160p and closer to 20 to 30 percent below 720p. Hardware HEVC decoding has shipped in Apple devices since 2015 and in virtually every 4K TV. Chrome has supported it since version 107 (2022) wherever the GPU provides a hardware decoder.

The less obvious part is the economics. A second codec ladder means more encoding compute, more storage and a split cache footprint. Whether HEVC pays for itself depends far more on your per-TB delivery rate and your catalog-to-traffic ratio than on the codec itself.

HEVC vs H264 compression efficiency: how much bitrate it actually saves

HEVC gets its efficiency from structural changes rather than tuning:

• Larger coding units. Coding tree units go up to 64x64 pixels, against H.264's 16x16 macroblocks.
• Finer intra prediction. HEVC has 35 intra prediction modes, against 9 for 4x4 blocks in H.264.
• Better motion signaling. Advanced motion vector prediction and merge mode cut the bits spent on motion data.
• An extra in-loop filter. Sample adaptive offset (SAO) runs after deblocking.

Large flat areas and slow motion benefit most. That is why the gain grows with resolution: a 4K frame has more of the redundancy that big blocks exploit.

In practice, typical figures for well-tuned software encoders look like this. A 1080p rendition that needs 5 to 6 Mb/s in H.264 lands around 3 to 4 Mb/s in HEVC. A 2160p rendition that needs 25 to 40 Mb/s in H.264 lands around 12 to 20 Mb/s in HEVC. These are labeled ranges, not guarantees: content complexity moves them by tens of percent.

Hardware encoders (GPU and ASIC) narrow the HEVC advantage, because they skip the expensive rate-distortion searches where much of the gain lives. For VOD, a useful starting point comes from the libx265 wrapper documentation: its default CRF of 28 is meant to look roughly like libx264 at CRF 23. Treat that as a calibration point only. Confirm every rung with VMAF or a comparable perceptual metric on your own titles before you trust a percentage.

H264 vs H265 at a glance: quality, bitrate, licensing and devices

HEVC wins every efficiency row and loses only on reach, encode cost and licensing clarity, which is why the dual-ladder setup remains the default in 2026.

Device support for H.265 streaming in 2026

H.265 streaming works natively on Apple platforms, smart TVs, streaming sticks and most Android phones. On desktop browsers it depends on a hardware decoder: Chrome and Edge play HEVC only when the GPU exposes one, and Firefox support varies by operating system. Any web audience therefore needs a runtime capability check and an H.264 fallback.

The hardware milestones that define your reachable base:

• Apple. A9 chips (2015) added hardware HEVC decode, and iOS 11 and macOS High Sierra (2017) enabled HEVC in HLS. Apple players expect fMP4 segments and the hvc1 tag; hev1-tagged streams can fail to play.
• Android. The platform has included an HEVC decoder since Android 5.0, and SoC hardware decode became common from about 2015. 10-bit Main10 on low-end devices is the usual gap.
• Desktop GPUs. Intel 6th-generation Core (2015) decodes 8-bit HEVC in hardware, with 10-bit from the 7th generation. NVIDIA added full HEVC decode with the second-generation Maxwell cards the same year.
• TVs and sticks. HEVC is the baseline codec for 4K and HDR10, so essentially every 4K panel and 4K streaming device decodes it.

On the web, do not infer support from the user agent. Query MediaSource.isTypeSupported with the exact CODECS string, for example hvc1.2.4.L123.B0 for Main10 at level 4.1. Then call MediaCapabilities.decodingInfo and check the powerEfficient flag. A device that only decodes HEVC in software will drain the battery and drop frames, so treat it as H.264-only.

In HLS, list both codecs in the multivariant playlist with accurate CODECS attributes and let the player choose. In DASH, put each codec in its own adaptation set.

Why HEVC licensing shaped browser support

H.264 has one dominant patent pool, and that pool has permanently waived royalties for internet video delivered free to end users. HEVC licensing is fragmented across several pools, including Access Advance and Via LA, plus licensors outside any pool.

Streamed content has generally not been the expensive part. The device and software side has. Shipping a software HEVC decoder exposes a browser or OS vendor to multiple royalty claims. That is why Chrome and Firefox lean on hardware decoders the chip vendor already licensed, and why Windows sells HEVC playback as a separate extension.

For a streaming service, the practical exposure is usually limited to encoders you distribute and apps that bundle their own decoder. Review that with counsel rather than assuming either codec is royalty-free in every context.

Does HEVC pay for itself? Egress math for a dual ladder

This blog's worked example uses five assumptions:

• The service delivers 200 TB a month, all in H.264 today.
• Player telemetry shows 70 percent of watch time on devices with efficient HEVC decode.
• HEVC saves 35 percent at matched VMAF across the ladder.
• The catalog is 10,000 hours, and the H.264 ladder rungs sum to about 15 Mb/s.
• Storage costs $0.015 to $0.023 per GB-month (typical object storage).

Egress drops by 200 x 0.70 x 0.35 = 49 TB, to 151 TB a month. The new HEVC ladder adds about 4.4 GB per catalog hour (6.75 GB x 0.65), or roughly 44 TB stored. That storage costs about $660 to $1,010 a month, plus a one-time encode bill that grows with every new title.

At hyperscaler list rates of roughly $20 to $85 per TB (estimate), the 49 TB saved is worth $980 to $4,165 a month. HEVC pays back quickly.

At low metered rates, the math flips. As of October 2026, BlazingCDN charges from $5 per TB down to $2.50 per TB at volume on progressive tiers, so 200 TB costs $765 and 151 TB costs $593.50 (the first 100 TB for $415, plus 51 TB at $3.50). Current rates are on the BlazingCDN pricing page, and you can model your own ladder with the CDN cost calculator for HEVC and H.264 egress. That $171.50 monthly saving does not cover the second ladder's storage.

The break-even rule this example produces is simple: HEVC pays on egress when TB saved per month multiplied by your per-TB delivery rate exceeds the extra TB stored multiplied by your storage rate, plus amortized encoding. Large catalogs with long tails and cheap delivery fail that test. Small catalogs with heavy traffic and expensive delivery pass it easily.

This blog's contrarian take: on cheap delivery, the strongest case for HEVC is not the bill. It is quality of experience. At the same 4 Mb/s, an HEVC viewer gets the 1080p rung instead of 720p, and HDR effectively requires HEVC.

BlazingCDN delivers pre-encoded HLS, LL-HLS and DASH and does not transcode. Your pipeline produces both ladders, and the CDN serves them as ordinary segments.

Trade-offs and failure modes of running HEVC beside H.264

• Split cache footprint. Two ladders double the object count per title. Long-tail titles see lower edge hit ratios and more origin fetches. An origin shield absorbs much of this, but watch shield egress after the rollout.
• Live encode budget. Real-time software HEVC at 1080p60 needs several times the cores of H.264. Most live operators use hardware HEVC encoders and accept a smaller gain, often 20 to 30 percent (estimate), or keep live on H.264.
• Silent software decode. Without the powerEfficient check, laptops play HEVC on the CPU, frames drop, and your rebuffer metrics blame the network.
• Packaging errors. Common causes of "plays on Android, fails on iPhone" are hev1 instead of hvc1, MPEG-TS instead of fMP4, and a wrong level in the CODECS string.
• ABR mixing. Players should not switch codecs mid-session. Keep each codec in its own variant group, or switches will cause decoder resets.

Which codec to stream in 2026: recommendation matrix

HEVC is the right primary codec wherever the audience sits on TVs and native apps, while H.264 still wins for browser-heavy, long-tail and low-latency workloads.

AV1 is the next comparison for web-heavy audiences and deserves its own evaluation rather than a footnote here.

FAQ: H.265 vs H.264 for streaming

H265 vs H264: which looks better at the same bitrate?

HEVC looks better at the same bitrate, typically matching H.264 quality with 30 to 50 percent fewer bits. The advantage is largest at 4K and in flat, slow-moving scenes, and smallest below 720p or with hardware encoders. Verify on your own content with VMAF at each ladder rung, because encoder settings and content complexity can shift the gain by tens of percent.

Is H.265 streaming supported in Chrome, Firefox and Safari?

Safari plays H.265 natively on Apple devices with A9 or later chips. Chrome has supported HEVC since version 107 in 2022, but only where the GPU provides a hardware decoder, and Edge behaves similarly. Firefox support varies by operating system. Web players should check MediaSource.isTypeSupported and MediaCapabilities.decodingInfo before selecting HEVC renditions.

Do I need to pay HEVC royalties to stream H.265 video?

Streaming H.265 content has generally not triggered royalty demands, but HEVC licensing is split across several patent pools and independent licensors. Exposure usually comes from distributing encoders or apps that bundle their own software decoder. H.264 is simpler, because its main pool waives royalties for internet video delivered free to viewers. Confirm your situation with counsel.

Should live streams use HEVC vs H264?

Most low-latency live streams still use H.264, because real-time software HEVC encoding needs several times the CPU per channel. Hardware HEVC encoders make live HEVC practical, though with smaller savings, often 20 to 30 percent (estimate). A common pattern is H.264 for browsers and mobile web, plus an HEVC ladder for TV and native app endpoints.

Does switching from H.264 to HEVC reduce CDN costs?

Switching to HEVC reduces delivered terabytes in proportion to the HEVC-capable share of viewing and the bitrate saving, often 20 to 35 percent of total egress. Net savings depend on your per-TB rate. Expensive delivery pays back quickly, while cheap metered delivery may not cover the storage and encoding cost of maintaining a second rendition ladder.

Run a dual-ladder break-even test on your own catalog

This week, encode your 20 most-watched titles in HEVC and match VMAF to your current H.264 rungs. Then sum segment bytes per rung to get your real saving. From player telemetry, pull the share of watch time on devices where decodingInfo reports powerEfficient HEVC.

Multiply three numbers: that share, the measured saving and your monthly terabytes. Compare the result, priced at your delivery rate, against the storage cost of the extra ladder. The answer tells you whether HEVC is a cost project or a quality project.

If you run the test with both ladders, the BlazingCDN HLS Streaming CDN delivers pre-encoded HLS, LL-HLS and DASH renditions of either codec during a 14-day testing period on real production traffic at $5 per TB, with the monthly minimum waived.