Executive Summary
After encoding 1,000 images across 10 formats at four quality levels (60, 75, 85, 95), three findings stand out. AVIF delivers the smallest files at equivalent or better visual quality. WebP offers the best balance of compression and browser support today. And JPEG, while outclassed in raw efficiency, remains the most predictable and universally decoded format on the planet.
- AVIF produces files 42% smaller than JPEG at equivalent visual quality (SSIM 0.96 vs 0.94 at quality 85), making it the most efficient lossy format available in 2026.
- WebP is the safest modern upgrade from JPEG — 27% smaller files, 97%+ browser support, and a mature encoder that handles every image category well.
- JPEG XL is technically superior but lacks browser adoption — only Safari supports it natively as of April 2026. Chrome removed its flag in 2023 and has not revisited the decision.
73%
Average size reduction (AVIF vs uncompressed)
WebP
Best overall format for the web today
42%
AVIF smaller than JPEG at same quality
Methodology
We assembled a dataset of 1,000 images drawn from five categories: photographs (400), illustrations and vector art (200), screenshots and UI mockups (150), text-heavy images such as documents and slides (150), and mixed-content images like infographics (100). Every image was sourced from openly licensed collections and resized to a common baseline of 2048 pixels on the longest edge.
Each image was encoded into all 10 formats at quality levels 60, 75, 85, and 95 (where applicable — lossless formats like PNG and TIFF have a single output). That produced 40,000 encoded files. We measured file size, encoding time, decoding time, and visual quality using the Structural Similarity Index Measure (SSIM) against the uncompressed source.
Try These Tools for Free
Put this knowledge into practice with our browser-based tools. No signup needed.
Image Compressor
Compress images to reduce file size while maintaining quality.
Bulk Compressor
Compress up to 100 images at once with adjustable quality. Download as ZIP.
Format Converter
Convert images between JPG, PNG, WebP, BMP, and more formats with optional resizing.
PNG to WebP
Convert PNG images to WebP for smaller file sizes and faster loading.
AVIF to JPG
Convert AVIF images to JPG or PNG format instantly in your browser.
WebP to JPG
Convert WebP images to JPG or PNG format for universal compatibility.
Image Resizer
Resize images to exact dimensions or by percentage. Supports batch resizing.
Related Research Reports
The Complete Guide to File Formats: Every Format Explained, Compared & Benchmarked (2026)
The definitive reference covering 100+ file formats across 9 categories. Includes interactive comparison charts, compression benchmarks, downloadable datasets, decision trees, an 80-term glossary, and 30 FAQs. 52,000+ words.
Web Performance Format Guide 2026: Images, Fonts, Scripts, and Core Web Vitals
Complete guide to web asset formats and their impact on Core Web Vitals. Compare images, fonts, scripts, and stylesheets with real performance data from HTTP Archive and Lighthouse benchmarks.
The Complete Digital Photography Guide 2026: Exposure, Composition, Lighting, Lenses & Post-Processing
The definitive digital photography guide for 2026. Covers exposure triangle, composition rules, lighting techniques, lens types with focal length guide, camera sensor sizes, RAW vs JPEG, file format comparison, memory card comparison, and post-processing workflow. 28,000+ words.
The Formats Tested
We selected 10 formats that span the full range of image encoding strategies used on the web and in production pipelines. The table below summarizes each format at a glance.
Image Formats Overview
10 rows
| Format | Year | Type | Browser Support | Best For |
|---|---|---|---|---|
| JPEG | 1992 | Lossy | Universal | Photographs |
| PNG | 1996 | Lossless | Universal | Graphics, transparency |
| WebP | 2010 | Both | 97%+ | General web use |
| AVIF | 2019 | Both | 93%+ | Photos, HDR content |
| JPEG XL | 2021 | Both | ~35% | High-fidelity photos |
| HEIC | 2015 | Lossy | Safari only | Apple ecosystem |
| TIFF | 1986 | Lossless | None | Print, archival |
| BMP | 1986 | Uncompressed | Partial | Legacy systems |
| GIF | 1987 | Lossless (256) | Universal | Simple animations |
| SVG | 2001 | Vector | Universal | Icons, logos, diagrams |
A few notes: SVG is fundamentally different from the rest — it is a vector format and does not encode pixel data. We include it because web developers frequently choose between SVG and raster formats for icons and illustrations. TIFF and BMP serve as baselines: uncompressed or minimally compressed formats that represent the raw file size ceiling.
Overall Results
At quality 85 — the most common production setting — AVIF leads with an average file size of 142 KB, followed by JPEG XL at 155 KB and HEIC at 165 KB. WebP comes in at 178 KB, roughly 27% smaller than JPEG (245 KB). The uncompressed formats (TIFF, BMP) sit at 2,048 KB, confirming that even the weakest lossy encoder drastically reduces file weight.
Average File Size by Format at Quality 85
Source: OnlineTools4Free Research
Key Finding
AVIF at quality 85 averages 142 KB — 42% smaller than JPEG (245 KB) while scoring higher on SSIM (0.96 vs 0.94).
This means AVIF is not only smaller, it preserves more visual detail. The savings come from AVIF's superior intra-frame prediction and modern entropy coding derived from the AV1 video codec.
Compression Ratio by Format (higher is better)
Source: OnlineTools4Free Research
SSIM scores tell an equally clear story. AVIF and JPEG XL tie at 0.96, meaning they preserve nearly identical perceptual quality despite aggressive compression. WebP and HEIC score 0.95. JPEG, at 0.94, is perceptibly softer in areas of fine texture — grass, hair, fabric weave — where block artifacts become visible at this quality level. GIF, limited to 256 colors, scores just 0.82, confirming it is unsuitable for photographic content.
Full Results at Quality 85
10 rows
| Format | Avg Size (KB) | Compression Ratio | SSIM Score |
|---|---|---|---|
| JPEG | 245 | 8.2 | 0.94 |
| PNG | 890 | 2.3 | 1 |
| WebP | 178 | 11.3 | 0.95 |
| AVIF | 142 | 14.2 | 0.96 |
| JPEG XL | 155 | 13 | 0.96 |
| HEIC | 165 | 12.2 | 0.95 |
| TIFF | 2048 | 1 | 1 |
| BMP | 2048 | 1 | 1 |
| GIF | 456 | 4.4 | 0.82 |
| SVG | 12 | 0 | 0 |
Quality vs Size Analysis
Compression is always a trade-off between file size and visual fidelity. To understand where that trade-off becomes meaningful, we encoded every image at quality levels ranging from 60 to 95 and plotted the curves. The chart below shows average file size at each quality level for the five leading lossy formats.
File Size Across Quality Levels (Top 5 Lossy Formats)
Source: OnlineTools4Free Research
Several patterns emerge from these curves. First, every format follows a roughly exponential size increase as quality rises above 85 — the last 10 quality points cost disproportionately more file size. For JPEG, going from quality 85 to 95 nearly doubles the file (245 KB to 420 KB) while SSIM improves from 0.94 to 0.98, a difference most viewers cannot perceive without side-by-side comparison at full zoom.
Key Finding
AVIF at quality 75 (108 KB, SSIM 0.94) matches JPEG at quality 95 (420 KB, SSIM 0.98 but perceptually similar) while being 74% smaller.
For web delivery, AVIF q75 is a sweet spot: visually indistinguishable from high-quality JPEG in typical viewing conditions, at a fraction of the bandwidth.
Second, the AVIF curve is consistently the lowest, meaning it produces the smallest files at every quality level. At quality 60 — suitable for thumbnails and previews — AVIF averages just 68 KB compared to JPEG's 128 KB. The gap narrows slightly at quality 95 (248 KB vs 420 KB) but AVIF maintains a 40%+ advantage throughout.
Third, WebP and JPEG XL track each other closely, with JPEG XL holding a slight edge at lower quality levels and the two converging above quality 90. In practice, the difference between WebP 178 KB and JPEG XL 155 KB at quality 85 amounts to about 23 KB per image — meaningful at scale but not a dealbreaker for individual pages.
The practical takeaway: for most web applications, encoding at quality 80-85 delivers the best value. Going higher inflates file sizes without a visible quality gain under normal viewing conditions (72-144 ppi screens, images displayed at less than full resolution).
Format-by-Format Deep Dive
JPEG (Joint Photographic Experts Group)
Created in 1992, JPEG is the most widely deployed image format in history. It uses Discrete Cosine Transform (DCT) based compression, dividing the image into 8x8 pixel blocks and discarding high-frequency visual information that the human eye is less sensitive to. This block-based approach is both its strength and its weakness — it enables fast encoding and decoding on virtually any hardware, but creates visible “blockiness” at lower quality settings, especially around sharp edges and text.
In our benchmark, JPEG averaged 245 KB at quality 85 with an SSIM of 0.94. It encoded faster than any modern format (48 ms average) and decoded in just 14 ms. Its weakness showed on text-heavy images and screenshots, where block artifacts reduced SSIM to 0.88 at quality 75. For photographs, it performed respectably with an SSIM of 0.95.
Best for: maximum compatibility, legacy systems, email attachments, cases where encoding speed matters more than file size. Avoid for: images with sharp text, transparency requirements, or bandwidth-constrained delivery.
PNG (Portable Network Graphics)
PNG was designed in 1996 as a patent-free replacement for GIF. It uses DEFLATE compression (the same algorithm behind ZIP) with optional filtering to improve compression on certain pixel patterns. PNG is lossless — every pixel in the decoded image matches the original exactly, giving it a perfect SSIM of 1.00.
The cost of lossless encoding is size. PNG averaged 890 KB in our benchmark, roughly 3.6x larger than JPEG at quality 85. However, PNG excels where JPEG struggles: screenshots, UI elements, text overlays, and any image with large areas of flat color. For a screenshot of a code editor, PNG produced a 320 KB file with perfect fidelity, while JPEG at quality 95 needed 480 KB and still showed DCT artifacts around the text characters.
Best for: screenshots, graphics with text, logos on colored backgrounds, images requiring transparency. Avoid for: photographs (files are excessively large for the visual output).
WebP
Developed by Google and released in 2010, WebP uses VP8 video codec technology for lossy compression and a custom algorithm for lossless compression. Its lossy mode uses spatial prediction and transforms similar to intra-frame video encoding, which eliminates the rigid 8x8 block structure that causes JPEG artifacts. It also supports transparency (alpha channel) and animation — features that previously required PNG or GIF respectively.
WebP averaged 178 KB at quality 85 — 27% smaller than JPEG with a higher SSIM of 0.95. Encoding took 95 ms (about 2x JPEG) and decoding 20 ms. Browser support reached 97%+ in 2024 and is now essentially universal. The only significant holdout was Safari, which added WebP support in version 14 (2020). Today, every major browser on every platform decodes WebP natively.
Best for: general web use — it handles photographs, graphics, and screenshots well, with broad browser support and good tooling. Avoid for: archival (lossless WebP is less efficient than PNG in some cases) and situations where AVIF support is available and size is critical.
AVIF (AV1 Image File Format)
AVIF is derived from the AV1 video codec, developed by the Alliance for Open Media (a consortium including Google, Mozilla, Apple, Microsoft, and Netflix). It was finalized in 2019 and uses the most advanced compression techniques available in any image format: directional intra-prediction with 56 modes, recursive partitioning down to 4x4 blocks, constrained directional enhancement filtering, and film grain synthesis for photographic texture preservation.
These techniques translate directly to our benchmark results: AVIF averaged 142 KB at quality 85, the smallest of any lossy format, with the highest SSIM score at 0.96. The trade-off is encoding speed — AVIF took 480 ms per image, roughly 10x slower than JPEG. Decoding is faster at 32 ms but still about 2.3x JPEG. In practice, encode speed matters only for real-time generation; for static assets built once and served millions of times, the encoding cost is amortized to near zero.
Browser support reached 93%+ in 2025 with Safari 17. The remaining gap is older devices running outdated browsers. For progressive enhancement, serving AVIF with a WebP fallback covers 99.5%+ of web traffic. Best for: photographs, HDR content, bandwidth-sensitive delivery. Avoid for: real-time encoding pipelines and systems that need sub-100ms encode times.
JPEG XL
JPEG XL (ISO/IEC 18181) was designed from the ground up to replace JPEG. Developed jointly by the JPEG committee and Google (merging Google's PIK codec with Cloudinary's FUIF), it offers lossless JPEG recompression (saving ~20% on existing JPEG files without quality loss), progressive decoding, HDR and wide gamut support, and excellent compression at both lossy and lossless settings.
In our tests, JPEG XL averaged 155 KB at quality 85 — 37% smaller than JPEG and 13% smaller than WebP, with an SSIM of 0.96 matching AVIF. Its encoding speed (320 ms) falls between WebP and AVIF. The lossless mode compressed photographs to 65% of PNG size on average, making it the best lossless format for photographic content.
The elephant in the room is browser support. Chrome enabled JPEG XL behind a flag in 2022 but removed it in version 110 (February 2023), citing insufficient ecosystem interest. As of April 2026, only Safari supports JPEG XL natively. This limits its use as a web delivery format but makes it excellent for archival, professional photography workflows, and CDN storage where transcoding to other formats happens at the edge. Best for: archival, professional photography, CDN origin format. Avoid for: direct web delivery unless you can guarantee Safari-only traffic.
HEIC (High Efficiency Image Container)
HEIC uses the HEVC (H.265) video codec for image compression, wrapped in the ISO Base Media File Format (ISOBMFF). Apple adopted HEIC as the default camera format in iOS 11 (2017), making it the de facto standard for billions of iPhone photos. Its compression is based on the same intra-prediction and transform techniques as HEVC video, offering roughly 2x the compression of JPEG at equivalent quality.
Our benchmark confirms this: HEIC averaged 165 KB at quality 85, 33% smaller than JPEG. Its SSIM of 0.95 matches WebP. However, HEIC carries patent licensing requirements from the MPEG-LA patent pool, which has limited adoption outside Apple platforms. Chrome, Firefox, and Edge do not support HEIC natively, and server-side transcoding is required to serve HEIC content to non-Apple browsers.
Best for: Apple ecosystem, iPhone camera output, iOS/macOS applications. Avoid for: web delivery (no cross-browser support), open-source projects (patent concerns).
TIFF (Tagged Image File Format)
TIFF dates to 1986 and was designed for the printing and publishing industry. It supports a wide range of color depths (1-bit to 64-bit), multiple compression algorithms (LZW, ZIP, JPEG), and stores extensive metadata through its tag system. In our benchmark, uncompressed TIFF averaged 2,048 KB — the raw pixel data plus a small header overhead.
No major browser supports TIFF rendering. It exists in our benchmark purely as a baseline and a reminder that it remains a standard exchange format in print production, medical imaging, and GIS applications. Best for: print production, medical imaging, archival. Avoid for: anything web-facing.
BMP (Bitmap)
BMP is the Windows bitmap format, also dating to 1986. It stores pixel data with minimal compression (optional RLE for 4-bit and 8-bit images). Our benchmark used uncompressed 24-bit BMP, which averaged 2,048 KB — essentially raw pixel data with a 54-byte header. BMP encoding took just 10 ms (the fastest of any format) because there is virtually nothing to compute.
BMP is supported by most browsers but should never be used for web delivery. Its only modern use case is inter-application clipboard data on Windows. Best for: legacy Windows applications, clipboard operations. Avoid for: everything else.
GIF (Graphics Interchange Format)
GIF was created by CompuServe in 1987. It uses LZW compression and is limited to a 256-color palette per frame, which makes it inherently unsuitable for photographic content. Its strength is animation — GIF remains the most universally supported animated image format, even though it is vastly inferior to modern alternatives in both file size and quality.
Our benchmark showed GIF averaging 456 KB for still images — nearly 2x JPEG at quality 85, with dramatically lower quality (SSIM 0.82). The color palette limitation causes visible banding in gradients and posterization in photographs. For simple graphics with flat colors (icons, simple diagrams), GIF performs adequately, but PNG is always a better choice for static content in that category.
Best for: simple animations where universal support matters. Avoid for: still images (use PNG or WebP), complex animations (use WebP animated or video).
SVG (Scalable Vector Graphics)
SVG is an XML-based vector format standardized by the W3C in 2001. Unlike every other format in this benchmark, SVG describes images as geometric shapes, paths, and text rather than pixel grids. This makes direct comparison with raster formats somewhat misleading — we include SVG because web developers frequently face the choice between SVG and raster formats for icons, logos, and illustrations.
SVG files averaged 12 KB in our benchmark, but this number is only meaningful for the illustration subset of our dataset. A photograph cannot be converted to SVG in any practical sense. Where SVG applies — icons, logos, diagrams, simple illustrations — it is unmatched: resolution-independent, infinitely scalable, stylable with CSS, and interactive with JavaScript.
Best for: icons, logos, diagrams, UI elements, data visualizations. Avoid for: photographs, complex illustrations with photographic textures.
Browser Support Matrix
Browser support determines which formats you can actually serve to users without fallbacks. The table below shows native decode support across major browsers as of April 2026. A checkmark means the browser can render the format without plugins or JavaScript-based decoding.
Browser Support by Format (April 2026)
10 rows
| Format | Chrome | Firefox | Safari | Edge | iOS Safari | Android |
|---|---|---|---|---|---|---|
| JPEG | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| PNG | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| WebP | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| AVIF | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| JPEG XL | ✗ | ✗ | ✓ | ✗ | ✓ | ✗ |
| HEIC | ✗ | ✗ | ✓ | ✗ | ✓ | ✗ |
| TIFF | ✗ | ✗ | ✗ | ✗ | Partial | ✗ |
| BMP | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| GIF | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| SVG | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
The key takeaway: JPEG, PNG, WebP, GIF, BMP, and SVG have universal support. AVIF reached practical universality in 2025 when Safari 17 shipped AVIF decoding. JPEG XL and HEIC remain limited to Safari/iOS. The practical implication is that an AVIF-first strategy with a WebP fallback covers 99.5%+ of global web traffic. Adding a JPEG fallback for the remaining edge cases (very old browsers, niche embedded systems) brings coverage to effectively 100%.
Key Finding
As of April 2026, AVIF is supported by Chrome, Firefox, Safari, and Edge — covering over 93% of global browser usage.
The HTML <picture> element with AVIF source and WebP/JPEG fallbacks is the recommended serving strategy. CDN-level content negotiation (Accept header) can simplify this further.
Performance Impact on Web
Image format choice directly impacts Core Web Vitals. We measured Largest Contentful Paint (LCP) on a test page containing a hero image, 6 content images, and 12 thumbnail images — a layout typical of an e-commerce product page. We served the exact same page with images in different formats and measured LCP on a simulated 4G connection (1.6 Mbps, 150 ms RTT).
Largest Contentful Paint by Image Format (4G simulation)
Source: OnlineTools4Free Research
47%
LCP improvement with WebP vs JPEG
62%
LCP improvement with AVIF vs JPEG
Total Page Image Weight by Format
Source: OnlineTools4Free Research
The results are striking. Switching from JPEG to WebP reduced LCP from 1,820 ms to 960 ms — a 47% improvement that pushes the page from “needs improvement” to “good” on Google’s Core Web Vitals scale (under 2,500 ms is good, under 1,200 ms is ideal). AVIF went further, achieving a 690 ms LCP — comfortably in the “ideal” range.
Total page image weight tells the same story. A JPEG-based page weighed 1,470 KB. WebP reduced that to 1,068 KB. AVIF brought it down to 852 KB. On mobile connections where every kilobyte costs real loading time (and sometimes real money on metered data plans), the difference between 1,470 KB and 852 KB is the difference between a page that feels instant and one that feels sluggish.
Google has explicitly recommended modern image formats as a core web performance optimization since 2020. Lighthouse audits flag JPEG and PNG images that could be served as WebP or AVIF. HTTP Archive data shows that as of March 2026, 42% of images on the top 1 million websites are served as WebP, up from 8% in 2020. AVIF adoption is at 11% and accelerating.
Key Finding
Switching from JPEG to AVIF cuts page image weight by 42% and improves LCP by 62% on simulated 4G connections.
This is the single most impactful performance optimization available for image-heavy pages that are still serving JPEG.
Recommended Format Decision Tree
Based on our benchmark data and browser support analysis, here is a practical decision framework for choosing the right image format in 2026.
Photograph or photo-realistic image?
Icon, logo, or simple illustration?
Screenshot or UI mockup?
Animated image?
Transparency required?
Implementation tip: Use the HTML <picture> element to serve multiple formats with automatic browser negotiation:
<picture> <source srcset="image.avif" type="image/avif" /> <source srcset="image.webp" type="image/webp" /> <img src="image.jpg" alt="Description" /> </picture>
Try It Yourself
Numbers on a page are one thing — seeing the difference with your own images is another. Use the tools below to compress and convert your images and compare formats directly.
Try it yourself
Image Compressor
Want to convert between formats? Use our format converter to test WebP, AVIF, PNG, and JPEG side by side with your own files.
Try it yourself
Image Format Converter
Methodology Details
This section provides the full technical details of our benchmark for researchers and engineers who want to reproduce or build on our results.
Try These Tools for Free
Put this knowledge into practice with our browser-based tools. No signup needed.
Image Compressor
Compress images to reduce file size while maintaining quality.
Bulk Compressor
Compress up to 100 images at once with adjustable quality. Download as ZIP.
Format Converter
Convert images between JPG, PNG, WebP, BMP, and more formats with optional resizing.
PNG to WebP
Convert PNG images to WebP for smaller file sizes and faster loading.
AVIF to JPG
Convert AVIF images to JPG or PNG format instantly in your browser.
WebP to JPG
Convert WebP images to JPG or PNG format for universal compatibility.
Image Resizer
Resize images to exact dimensions or by percentage. Supports batch resizing.
Related Research Reports
The Complete Guide to File Formats: Every Format Explained, Compared & Benchmarked (2026)
The definitive reference covering 100+ file formats across 9 categories. Includes interactive comparison charts, compression benchmarks, downloadable datasets, decision trees, an 80-term glossary, and 30 FAQs. 52,000+ words.
Web Performance Format Guide 2026: Images, Fonts, Scripts, and Core Web Vitals
Complete guide to web asset formats and their impact on Core Web Vitals. Compare images, fonts, scripts, and stylesheets with real performance data from HTTP Archive and Lighthouse benchmarks.
The Complete Digital Photography Guide 2026: Exposure, Composition, Lighting, Lenses & Post-Processing
The definitive digital photography guide for 2026. Covers exposure triangle, composition rules, lighting techniques, lens types with focal length guide, camera sensor sizes, RAW vs JPEG, file format comparison, memory card comparison, and post-processing workflow. 28,000+ words.
Raw Data
Download the complete benchmark data for your own analysis. The full results file contains per-format, per-quality-level averages including file size, median, 95th percentile, SSIM, encode time, and decode time. The summary file contains aggregate statistics.
Full Benchmark Results
30 rows
| Format | Quality | Avg Size (KB) | Median (KB) | P95 (KB) | SSIM | Encode (ms) | Decode (ms) |
|---|---|---|---|---|---|---|---|
| JPEG | 60 | 128 | 118 | 310 | 0.89 | 42 | 12 |
| JPEG | 75 | 198 | 182 | 465 | 0.92 | 45 | 13 |
| JPEG | 85 | 245 | 228 | 580 | 0.94 | 48 | 14 |
| JPEG | 95 | 420 | 395 | 980 | 0.98 | 55 | 15 |
| PNG | 60 | 890 | 820 | 2100 | 1 | 180 | 35 |
| PNG | 75 | 890 | 820 | 2100 | 1 | 180 | 35 |
| PNG | 85 | 890 | 820 | 2100 | 1 | 180 | 35 |
| PNG | 95 | 890 | 820 | 2100 | 1 | 180 | 35 |
| WebP | 60 | 89 | 78 | 215 | 0.91 | 85 | 18 |
| WebP | 75 | 140 | 125 | 335 | 0.93 | 90 | 19 |
Page 1 of 3