How this report is built — read this before the tables
The quality scores are not ours. Every ranked model in the catalog is matched by name to an entry on Artificial Analysis, an independent benchmarking site that runs public human-preference arenas for image, video and speech models, and each row carries that arena's current ELO for the matched entry. The tables below show the exact Artificial Analysis leaderboard name each score was taken from, and each section links the specific arena. The credit prices are ours; the scores are theirs.
One honest caveat, stated here rather than in a footnote: the catalog stores each model's current arena score only — there is no score history. Every “frontier over time” view on this page is therefore constructed by grouping models by their release date and reading today's scores across those groups. It shows what each generation of models is worth now, not what any leaderboard looked like on a past date — older models' scores keep moving as new arena votes come in.
Price means one thing throughout: credits per job — the fewest credits one complete generation can cost on Versely, taken from each model's own price matrix. That is the only figure comparable across models that bill per second, per megapixel or flat per call. Scored models that publish only a per-1,000-character rate are listed with that rate labeled as such and kept out of the charts and the frontier. The pool is the published model catalog — one row per model page, so every row here is a model you can actually open and run.
Reading the frontier
A model is on the frontier when nothing in its arena scores higher at its price or below. Everything off the frontier is dominated: some other model gives more score for the same credits or the same score for fewer. Frontier rows are highlighted in every table; cohorts with fewer than 3 priced models are shown in the cohort tables but not narrated.
What the data says
- Text-to-video: the 2025 cohort peaks at 959 ELO, the 2026 cohort at 1143. Outscoring the 2025 peak now starts at 4 credits per job (Wan 3.0 Text to Video).
- Image-to-video: the 2025 cohort peaks at 947 ELO, the 2026 cohort at 1024. Outscoring the 2025 peak now starts at 17 credits per job (Happy Horse 1.1 Image to Video).
- Text-to-image: the 2024 cohort peaks at 871 ELO, the 2026 cohort at 1189. Outscoring the 2024 peak now starts at 1 credits per job (FLUX 3 Image Klein).
- Image editing: the 2025 cohort peaks at 1119 ELO, the 2026 cohort at 1183. Outscoring the 2025 peak now starts at 4 credits per job (GPT Image 2.5 Flare Edit).
Text-to-video
18 scored models · scores from the Artificial Analysis Text to Video arena
The strongest 2025 release in this pool, VEO 3.1, holds 959 ELO today; a clip starts at 32 credits. The strongest 2026 release, HeyGen Video, holds 1143 ELO; a clip starts at 8.
The cheapest way to outscore the best 2025 release is Wan 3.0 Text to Video (1131 ELO, released 2026), where a clip starts at 4 credits.
| Released in | Priced models | Best score today | Lowest job floor |
|---|---|---|---|
| 2025 | 3 | 959 ELO — VEO 3.1 at 32-192 credits | 2-32 credits — PrunaAI P-Video at 801 ELO |
| 2026 | 15 | 1143 ELO — HeyGen Video at 8-72 credits | 4-120 credits — Wan 3.0 Text to Video at 1131 ELO |
Every scored text-to-video model, by arena score
| # | Model | Released | Credits per job | Arena ELO | Scored as (Artificial Analysis) |
|---|---|---|---|---|---|
| 1 | HeyGen VideoFrontier | September 30, 2026 | 8-72 credits | 1143 | HeyGen Video 1 |
| 2 | Wan 3.0 Text to VideoFrontier | August 24, 2026 | 4-120 credits | 1131 | Wan 3.0 |
| 3 | Minimax H3 Text to Video | August 5, 2026 | 12-78 credits | 1120 | MiniMax H3 (768p) |
| 4 | Seedance 2.5 | August 1, 2026 | 22-320 credits | 1103 | Dreamina Seedance 2.5 |
| 5 | Flux 3 Text to Video | August 5, 2026 | 34-136 credits | 1091 | FLUX 3 |
| 6 | Seedance 2.0 | February 12, 2026 | 22-1639 credits | 1085 | Dreamina Seedance 2.0 |
| 7 | Grok Imagine Video 1.5 Text to Video | May 1, 2026 | 4-150 credits | 1038 | Grok Imagine Video 1.5 |
| 8 | Happy Horse 1.1 Text to Video | June 23, 2026 | 17-108 credits | 1034 | HappyHorse-1.1 |
| 9 | Wan 2.7 Text to Video | April 6, 2026 | 8-90 credits | 1023 | Wan2.7-260612 |
| 10 | Kling Video V3 Pro Text to Video | February 5, 2026 | 34-101 credits | 1000 | Kling 3.0 1080p (Pro) |
| 11 | Pixverse V6 Text to Video | March 30, 2026 | 5-29 credits | 991 | PixVerse V6 |
| 12 | VEO 3.1 | October 15, 2025 | 32-192 credits | 959 | Veo 3.1 |
| 13 | LTX 2.5 Text to Video Pro | August 24, 2026 | 58-136 credits | 927 | LTX-2.5 Pro |
| 14 | Runway Gen-4.5 | December 1, 2025 | 200-400 credits | 910 | Runway Gen-4.5 (720p) |
| 15 | Vidu Q3 Video | January 30, 2026 | 23-67 credits | 902 | Vidu Q3 Pro |
| 16 | LTX 2.3 Text to Video Fast | March 5, 2026 | 20-256 credits | 848 | LTX-2.3 Fast |
| 17 | Boreal | September 15, 2026 | 4-192 credits | 826 | Boreal |
| 18 | PrunaAI P-VideoFrontier | July 29, 2025 | 2-32 credits | 801 | P-Video |
# is the model's position within Versely's published catalog for this arena, not its position on the full Artificial Analysis leaderboard.
Image-to-video
8 scored models · scores from the Artificial Analysis Image to Video arena
The strongest 2025 release in this pool, VEO First Last Frame, holds 947 ELO today; a clip starts at 32 credits. The strongest 2026 release, Happy Horse 1.1 Image to Video, holds 1024 ELO; a clip starts at 17.
The cheapest way to outscore the best 2025 release is Happy Horse 1.1 Image to Video (1024 ELO, released 2026), where a clip starts at 17 credits.
| Released in | Priced models | Best score today | Lowest job floor |
|---|---|---|---|
| 2025 | 4 | 947 ELO — VEO First Last Frame at 32-128 credits | 12-80 credits — Pika 2.5 Image to Video at 847 ELO |
| 2026 | 4 | 1024 ELO — Happy Horse 1.1 Image to Video at 17-108 credits | 17-108 credits — Happy Horse 1.1 Image to Video at 1024 ELO |
Every scored image-to-video model, by arena score
| # | Model | Released | Credits per job | Arena ELO | Scored as (Artificial Analysis) |
|---|---|---|---|---|---|
| 1 | Happy Horse 1.1 Image to VideoFrontier | June 23, 2026 | 17-108 credits | 1024 | HappyHorse-1.1 |
| 2 | Kling O3 Pro Image to Video | February 5, 2026 | 56-168 credits | 1015 | Kling 3.0 Omni 1080p (Pro) |
| 3 | VEO First Last Frame | October 15, 2025 | 32-128 credits | 947 | Veo 3.1 |
| 4 | Vidu Q3 Image to Video | January 30, 2026 | 23-67 credits | 906 | Vidu Q3 Pro |
| 5 | Runway Gen-4.5 | December 1, 2025 | 200-400 credits | 896 | Runway Gen-4.5 (720p) |
| 6 | Kling O1 Image to Video | December 1, 2025 | 45-90 credits | 865 | Kling O1 Pro (January) |
| 7 | Pika 2.5 Image to VideoFrontier | November 1, 2025 | 12-80 credits | 847 | Pika 2.5 |
| 8 | LTX 2.3 Image to Video Pro | March 5, 2026 | 29-192 credits | 812 | LTX-2.3 Pro |
# is the model's position within Versely's published catalog for this arena, not its position on the full Artificial Analysis leaderboard.
Text-to-image
34 scored models · scores from the Artificial Analysis Text to Image arena
The strongest 2024 release in this pool, Flux 1.1 Pro, holds 871 ELO today; an image starts at 4 credits. The strongest 2026 release, GPT Image 2 Text to Image, holds 1189 ELO; an image starts at 9.
The cheapest way to outscore the best 2024 release is FLUX 3 Image Klein (1043 ELO, released 2026), where an image starts at 1 credits.
| Released in | Priced models | Best score today | Lowest job floor |
|---|---|---|---|
| 2024 | 3 | 871 ELO — Flux 1.1 Pro at 4 credits | 1 credit — Luma Photon Flash at 844 ELO |
| 2025 | 12 | 1123 ELO — GPT Image 1.5 at 2 credits | 1 credit — Wan 2.6 Text to Image at 1015 ELO |
| 2026 | 19 | 1189 ELO — GPT Image 2 Text to Image at 9-17 credits | 1 credit — FLUX 3 Image Klein at 1043 ELO |
Every scored text-to-image model, by arena score
| # | Model | Released | Credits per job | Arena ELO | Scored as (Artificial Analysis) |
|---|---|---|---|---|---|
| 1 | GPT Image 2 Text to ImageFrontier | April 21, 2026 | 9-17 credits | 1189 | GPT Image 2 (high) |
| 2 | Nano Banana 2.1Frontier | October 8, 2026 | 4-7 credits | 1176 | Nano Banana 2.1 |
| 3 | Grok Imagine Image 2.0Frontier | August 24, 2026 | 2-3 credits | 1168 | Grok Imagine Image 2.0 |
| 4 | Nano Banana 2 | February 26, 2026 | 3-6 credits | 1140 | Nano Banana 2 (Gemini 3.1 Flash Image) |
| 5 | GPT Image 1.5 | December 16, 2025 | 2 credits | 1123 | GPT Image 1.5 (high) |
| 6 | FLUX 3 Image | October 1, 2026 | 4 credits | 1119 | FLUX 3 Image |
| 7 | Seedream 5.0 Pro | August 5, 2026 | 6 credits | 1101 | Seedream 5.0 Pro |
| 8 | Qwen Image 3 Text to Image | August 1, 2026 | 3 credits | 1072 | Qwen-Image-3.0 |
| 9 | Grok Imagine Image Quality | July 28, 2025 | 4 credits | 1049 | grok-imagine-image-quality |
| 10 | FLUX 3 Image KleinFrontier | October 15, 2026 | 1 credit | 1043 | FLUX 3 Image Klein |
| 11 | Fibo 2.0 | October 19, 2026 | 4 credits | 1040 | Fibo 2.0 |
| 12 | Seedream 4.5 | December 4, 2025 | 4 credits | 1027 | Seedream 4.5 |
| 13 | Flux 2 Flex | November 25, 2025 | 4 credits | 1026 | FLUX.2 [flex] |
| 14 | Luma UNI 1 Max | March 5, 2026 | 9 credits | 1024 | Luma UNI 1 Max |
| 15 | Grok Imagine Image | July 28, 2025 | 2 credits | 1022 | grok-imagine-image |
| 16 | Flux 2 Max | December 16, 2025 | 6 credits | 1021 | FLUX.2 [max] |
| 17 | HiDream O1 Image | May 8, 2026 | 1 credit | 1017 | HiDream-O1-Image-1.5 |
| 18 | Wan 2.6 Text to Image | December 16, 2025 | 1 credit | 1015 | Wan2.6 Text to Image |
| 19 | Krea 2 Medium | May 12, 2026 | 3 credits | 1012 | Krea 2 Medium |
| 20 | Ideogram V4 | June 3, 2026 | 5 credits | 1002 | Ideogram 4.0 |
| 21 | Cosmos 3 Super | June 1, 2026 | 4 credits | 993 | Cosmos3-Super-Text2Image (agentic) |
| 22 | Flux 2 Flash | January 15, 2026 | 1 credit | 988 | FLUX.2 [dev] Flash |
| 23 | Nano Banana | August 26, 2025 | 2 credits | 985 | Nano Banana (Gemini 2.5 Flash Image) |
| 24 | Qwen Image | August 4, 2025 | 1 credit | 953 | Qwen Image Plus 2601 |
| 25 | Z Image Turbo | November 27, 2025 | 1 credit | 936 | Z-Image Turbo |
| 26 | Flux 2 Klein 9B Base | January 15, 2026 | 1 credit | 930 | FLUX.2 [klein] 9B |
| 27 | Kling Image 3.0 | February 5, 2026 | 3 credits | 919 | Kling Image 3.0 Omni |
| 28 | ERNIE Image Turbo | April 15, 2026 | 1 credit | 915 | ERNIE Image Turbo |
| 29 | Qwen Z Image | November 27, 2025 | 2 credits | 891 | Z-Image Base |
| 30 | Leonardo Lucid Origin | August 5, 2025 | 4 credits | 878 | Lucid Origin Ultra |
| 31 | Flux 1.1 Pro | October 2, 2024 | 4 credits | 871 | FLUX1.1 [pro] |
| 32 | Luma Photon Flash | November 25, 2024 | 1 credit | 844 | Luma Photon |
| 33 | HiDream O1 Image Dev | May 8, 2026 | 1 credit | 809 | HiDream-O1-Image-Dev |
| 34 | Flux Schnell | August 1, 2024 | 1 credit | 740 | FLUX.1 [schnell] |
# is the model's position within Versely's published catalog for this arena, not its position on the full Artificial Analysis leaderboard.
Image editing
18 scored models · scores from the Artificial Analysis Image Editing arena
The strongest 2025 release in this pool, GPT Image 1.5, holds 1119 ELO today; an edit starts at 2 credits. The strongest 2026 release, GPT Image 2.5 Flare Edit, holds 1183 ELO; an edit starts at 4.
The cheapest way to outscore the best 2025 release is GPT Image 2.5 Flare Edit (1183 ELO, released 2026), where an edit starts at 4 credits.
| Released in | Priced models | Best score today | Lowest job floor |
|---|---|---|---|
| 2025 | 8 | 1119 ELO — GPT Image 1.5 at 2 credits | 1 credit — Qwen Image Edit 2511 at 993 ELO |
| 2026 | 10 | 1183 ELO — GPT Image 2.5 Flare Edit at 4-7 credits | 1 credit — Pruna Image Edit at 985 ELO |
Every scored image editing model, by arena score
| # | Model | Released | Credits per job | Arena ELO | Scored as (Artificial Analysis) |
|---|---|---|---|---|---|
| 1 | GPT Image 2.5 Flare EditFrontier | September 8, 2026 | 4-7 credits | 1183 | GPT Image 2.5 Flare (max) |
| 2 | Nano Banana 2.1 | October 8, 2026 | 4-7 credits | 1151 | Nano Banana 2.1 |
| 3 | Mai Image 2.5 Edit | June 2, 2026 | 4 credits | 1127 | MAI-Image-2.5 |
| 4 | FLUX 3 Image Edit | October 1, 2026 | 4 credits | 1123 | FLUX 3 Image |
| 5 | GPT Image 1.5Frontier | December 16, 2025 | 2 credits | 1119 | GPT Image 1.5 (high) |
| 6 | HunyuanImage 3.0 Instruct Edit | January 26, 2026 | 8 credits | 1064 | HunyuanImage 3.0 Instruct |
| 7 | Wan 2.7 Pro Edit | April 6, 2026 | 6 credits | 1051 | Wan 2.7 Pro |
| 8 | GPT Image 1 Edit | April 23, 2025 | 2 credits | 998 | GPT Image 1 (high) |
| 9 | Qwen Image Edit 2511Frontier | December 23, 2025 | 1 credit | 993 | Qwen Image Edit Plus 2511 |
| 10 | Pruna Image Edit | January 1, 2026 | 1 credit | 985 | P-Image-Edit |
| 11 | Qwen Image Edit | August 19, 2025 | 1 credit | 984 | Qwen Image Edit Max 2601 |
| 12 | Qwen Image 2 Edit | February 10, 2026 | 3 credits | 968 | Qwen Image 2.0 (2026-03-03) |
| 13 | HiDream O1 Image Edit | May 8, 2026 | 1 credit | 955 | HiDream-O1-Image |
| 14 | Kling Image O1 | December 1, 2025 | 3 credits | 944 | Kling Image O1 |
| 15 | Flux 2 Klein 4B Base Edit | January 15, 2026 | 1 credit | 941 | FLUX.2 [klein] 4B |
| 16 | Flux Kontext | May 29, 2025 | 1 credit | 915 | FLUX.1 Kontext [pro] |
| 17 | Seedream Edit | September 9, 2025 | 1 credit | 901 | SeedEdit 3.0 |
| 18 | LongCat Edit | December 5, 2025 | 12 credits | 897 | LongCat Image |
# is the model's position within Versely's published catalog for this arena, not its position on the full Artificial Analysis leaderboard.
Text-to-speech
11 scored models · scores from the Artificial Analysis Text to Speech arena
This arena's pool is too small to cut into release-year cohorts honestly — fewer than two cohorts reach 3 priced models — so it gets the full ranking only, no trend claims.
Every scored text-to-speech model, by arena score
| # | Model | Released | Credits per job | Arena ELO | Scored as (Artificial Analysis) |
|---|---|---|---|---|---|
| 1 | Eleven v4 | September 28, 2026 | 7 credits (headline rate) | 1334 | Eleven v4 |
| 2 | Cartesia Sonic 3.6 | August 27, 2026 | 4 credits (headline rate) | 1293 | Sonic 3.6 |
| 3 | Gemini 3.8 Flash TTS | September 23, 2026 | 2 credits (headline rate) | 1282 | Gemini 3.8 Flash TTS |
| 4 | Inworld TTS 2 | May 5, 2026 | 2 credits (headline rate) | 1266 | Realtime TTS-2 |
| 5 | Inworld TTS 2 Flash | September 2, 2026 | 2 credits (headline rate) | 1215 | Realtime TTS-2 Flash |
| 6 | Gemini 3.1 Flash TTS | April 15, 2026 | 12 credits (headline rate) | 1213 | Gemini 3.1 Flash TTS |
| 7 | Cartesia Sonic 3.5 | June 16, 2026 | 4 credits (headline rate) | 1193 | Sonic 3.5 |
| 8 | MiniMax Speech | April 2, 2025 | 2 credits (headline rate) | 1173 | Speech 2.8 HD |
| 9 | Grok TTS | March 16, 2026 | 4 credits (headline rate) | 1157 | SpaceXAI TTS |
| 10 | Chatterbox TTS | May 1, 2025 | 2 credits (headline rate) | 1108 | Chatterbox HD |
| 11 | ElevenLabs Speech Turbo | July 19, 2024 | 4 credits (headline rate) | 1092 | Turbo v2.5 |
# is the model's position within Versely's published catalog for this arena, not its position on the full Artificial Analysis leaderboard. 11 of the 11 rows publish no complete-job minimum — their credit figure is the headline rate, labeled, and they are excluded from the chart and the frontier.
Turn the frontier into a pick
This page tells you where quality stops being worth extra credits — it does not know your aspect ratio, clip length or style. The /best rankings cut the same catalog by use case (platform, resolution, budget, speed), and every model in every table above links to its page in the model catalog with full pricing, settings and prompt guidance. New releases land in what's new first — when the frontier moves, it moves there.
Frequently asked questions
Where do the quality scores on this page come from?+
From Artificial Analysis (artificialanalysis.ai), an independent AI benchmarking site that runs public human-preference arenas for text-to-image, image editing, text-to-video, image-to-video and text-to-speech models. Versely syncs those arena ELO scores into its catalog and matches them to catalog models by name — the 'scored as' column in each table shows the exact Artificial Analysis leaderboard entry a score was taken from. The scores are theirs; the credit prices are ours.
What does 'credits per job' mean?+
The fewest credits one complete generation can cost on Versely — the price matrix's own minimum for a full job, not a per-second or per-character rate read out of context. It is the only credit figure that is comparable across models that bill per second, per megapixel, or flat per call. A handful of scored audio models publish only a per-1,000-character rate with no complete-job minimum; those rows are labeled 'headline rate' and are excluded from the charts and the frontier rather than passed off as job prices.
Is the 'frontier over time' a history of past leaderboards?+
No, and the page says so where the numbers are used: the catalog stores each model's current arena score only. The cohort views group models by their release year and read today's scores across those groups. That shows what each generation of models is worth now — it does not show what the leaderboard looked like on any past date, because scores of older models keep moving as new arena votes come in.
Does a higher arena ELO mean that model is better for my job?+
Not automatically. Arena ELO measures which model's output people preferred across broad prompt sets — it says nothing about your specific style, aspect ratio, clip length or budget. Use this report to find the price band where quality stops improving for you, then check the use-case rankings under /best and the individual model pages under /models for the constraints that actually decide the pick.
Run the frontier models inside Versely
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync, in your browser or on your phone.
One account for the web and iPhone.