SenseNova U1.5 Litevs.Chroma1-HD
Which one is better?
In our own test, SenseNova U1.5 Lite leads. Neither is in the arena. Neither has an API price.
Test images
Same task, the first image of each. The task with the biggest difference is preselected.
Result for “Product photography”:SenseNova U1.5 Lite ahead
What the rating looks at: Materials, reflections, and fine details.
“Ahead”: better in all three rating rounds. “Slightly ahead”: a higher value, but not in every round.
Test protocol of the images
Each test attempt uses the same archived prompt and one request. The first output of that attempt counts. Repeats after technical test failures are disclosed for the affected images. Files are normalized to 1,024 × 1,024 pixels without cropping. Provider resolution, randomness, quality tier and prompt processing can still differ. The images are a practical visual test, not a reconstruction of the arena configuration.
Character and detail
SenseNova U1.5 Lite: manual, local stable-diffusion.cpp@3be23788fc0c9f35a2c24b1211caae11d4a54dbc, ggml@d49c49e1dd361ba3c0a77c90b5b1404b83969aa8; sensenova/SenseNova-U1.5-8B-MoT@9feeeab8a2792514d109cd34589342a2cc1d4ab2; native 2048x2048, 1:1; one request, first output; steps=50, guidance=4.0, seed=8087247682796819534; Apple M5 Max Metal MTL0 requested through the native backend option; automatic placement remains enabled, and actual placement is recorded in the CLI log; Original BF16 checkpoint. The exact two SenseNova vision convolution weights use a targeted lossless BF16-to-F32 rule; other eligible weights use the native unquantized f16 option. This BF16-to-F16 conversion rounds values, as quantified by the linked complete-checkpoint GGML preflight; excluded weights remain BF16 if used. Operation precision follows the backend.; Euler, flow shift 3; CPU Philox RNG (CLI cuda RNG option, no CUDA hardware); original BF16 weights without requested quantization; observational latent and pre-clipping finite checks; native prompt template without rewriting; empty unconditional prompt, source 2048x2048, stored 1024x1024 (Lanczos3 downscale), generated Nov 2, 2026, 10:17 PM. Source
Chroma1-HD: manual, local PyTorch 2.13.0, Diffusers 0.36.0; lodestones/Chroma1-HD@0e0c60ece1e82b17cb7f77342d765ba5024c40c0; native 1024x1024, 1:1; one request, first output; steps=40, guidance=3.0, seed=4845693986590311438; mps for text encoder and denoising; cpu for native VAE decode; bfloat16 for text encoder and transformer; float32 for VAE, generated Nov 2, 2026, 7:16 PM. Source
Protocol gradually-image-comparison-v1. Prompt SHA 256: 985a75b38642a9f0e5e0c0e0c67f1b7f67780d42bd2168fa24e9a7a7a94b0da6
SenseNova U1.5 Lite image SHA 256: c092b333f479c5b28ad33958d0e45f0fb0989970af02690079cf1398742c55b8
Chroma1-HD image SHA 256: 7ca51c510884638a4d8c09d0e361e2bf0328c466d36ec4235f5305580dc10a8a
Infographic
SenseNova U1.5 Lite: manual, local stable-diffusion.cpp@3be23788fc0c9f35a2c24b1211caae11d4a54dbc, ggml@d49c49e1dd361ba3c0a77c90b5b1404b83969aa8; sensenova/SenseNova-U1.5-8B-MoT@9feeeab8a2792514d109cd34589342a2cc1d4ab2; native 2048x2048, 1:1; one request, first output; steps=50, guidance=4.0, seed=2640628636372692680; Apple M5 Max Metal MTL0 requested through the native backend option; automatic placement remains enabled, and actual placement is recorded in the CLI log; Original BF16 checkpoint. The exact two SenseNova vision convolution weights use a targeted lossless BF16-to-F32 rule; other eligible weights use the native unquantized f16 option. This BF16-to-F16 conversion rounds values, as quantified by the linked complete-checkpoint GGML preflight; excluded weights remain BF16 if used. Operation precision follows the backend.; Euler, flow shift 3; CPU Philox RNG (CLI cuda RNG option, no CUDA hardware); original BF16 weights without requested quantization; observational latent and pre-clipping finite checks; native prompt template without rewriting; empty unconditional prompt, source 2048x2048, stored 1024x1024 (Lanczos3 downscale), generated Nov 2, 2026, 10:25 PM. Source
Chroma1-HD: manual, local PyTorch 2.13.0, Diffusers 0.36.0; lodestones/Chroma1-HD@0e0c60ece1e82b17cb7f77342d765ba5024c40c0; native 1024x1024, 1:1; one request, first output; steps=40, guidance=3.0, seed=3726635088695005336; mps for text encoder and denoising; cpu for native VAE decode; bfloat16 for text encoder and transformer; float32 for VAE, generated Nov 2, 2026, 7:22 PM. Source
Protocol gradually-image-comparison-v1. Prompt SHA 256: ed65cb4e6a51e2336f8b149ad4cbb16e020f07d527ad94362e719d2dcf91d94f
SenseNova U1.5 Lite image SHA 256: dcf8206386da05d11c35a6950309e61f08a504d440366226eac316328718f5ae
Chroma1-HD image SHA 256: 834ab94e5514295d56b66f606572d348fc4b5d1cdeba5874921c79c00774c76a
Product photography
SenseNova U1.5 Lite: manual, local stable-diffusion.cpp@3be23788fc0c9f35a2c24b1211caae11d4a54dbc, ggml@d49c49e1dd361ba3c0a77c90b5b1404b83969aa8; sensenova/SenseNova-U1.5-8B-MoT@9feeeab8a2792514d109cd34589342a2cc1d4ab2; native 2048x2048, 1:1; one request, first output; steps=50, guidance=4.0, seed=1950293105094404455; Apple M5 Max Metal MTL0 requested through the native backend option; automatic placement remains enabled, and actual placement is recorded in the CLI log; Original BF16 checkpoint. The exact two SenseNova vision convolution weights use a targeted lossless BF16-to-F32 rule; other eligible weights use the native unquantized f16 option. This BF16-to-F16 conversion rounds values, as quantified by the linked complete-checkpoint GGML preflight; excluded weights remain BF16 if used. Operation precision follows the backend.; Euler, flow shift 3; CPU Philox RNG (CLI cuda RNG option, no CUDA hardware); original BF16 weights without requested quantization; observational latent and pre-clipping finite checks; native prompt template without rewriting; empty unconditional prompt, source 2048x2048, stored 1024x1024 (Lanczos3 downscale), generated Nov 2, 2026, 10:11 PM. Source
Chroma1-HD: manual, local PyTorch 2.13.0, Diffusers 0.36.0; lodestones/Chroma1-HD@0e0c60ece1e82b17cb7f77342d765ba5024c40c0; native 1024x1024, 1:1; one request, first output; steps=40, guidance=3.0, seed=7965145008513162443; mps for text encoder and denoising; cpu for native VAE decode; bfloat16 for text encoder and transformer; float32 for VAE, generated Nov 2, 2026, 7:10 PM. Source
Protocol gradually-image-comparison-v1. Prompt SHA 256: c67fd71e91304458a2feef5c91efa8677260fab0c0099a29a2c52cb579f5b2f3
SenseNova U1.5 Lite image SHA 256: d425ed5236325ee28633f83cb00f3fd39181f7dd6df0e5408d8b1c07c6796107
Chroma1-HD image SHA 256: 8a0ebbb527f4db84a42a98a256333b00b4dc5ffae699a31b694f36bf34613e4b
Typography and layout
SenseNova U1.5 Lite: manual, local stable-diffusion.cpp@3be23788fc0c9f35a2c24b1211caae11d4a54dbc, ggml@d49c49e1dd361ba3c0a77c90b5b1404b83969aa8; sensenova/SenseNova-U1.5-8B-MoT@9feeeab8a2792514d109cd34589342a2cc1d4ab2; native 2048x2048, 1:1; one request, first output; steps=50, guidance=4.0, seed=8328647217678382438; Apple M5 Max Metal MTL0 requested through the native backend option; automatic placement remains enabled, and actual placement is recorded in the CLI log; Original BF16 checkpoint. The exact two SenseNova vision convolution weights use a targeted lossless BF16-to-F32 rule; other eligible weights use the native unquantized f16 option. This BF16-to-F16 conversion rounds values, as quantified by the linked complete-checkpoint GGML preflight; excluded weights remain BF16 if used. Operation precision follows the backend.; Euler, flow shift 3; CPU Philox RNG (CLI cuda RNG option, no CUDA hardware); original BF16 weights without requested quantization; observational latent and pre-clipping finite checks; native prompt template without rewriting; empty unconditional prompt; three local execution attempts: original Metal BF16 attempt aborted on a missing matrix kernel before output; same-seed CPU retry deliberately interrupted before its first sampling step without image or latent; first image retained after the reviewed loader compatibility correction and disclosed F16 runtime conversion with unchanged prompt and seed, source 2048x2048, stored 1024x1024 (Lanczos3 downscale), generated Nov 2, 2026, 10:04 PM. Source
Chroma1-HD: manual, local PyTorch 2.13.0, Diffusers 0.36.0; lodestones/Chroma1-HD@0e0c60ece1e82b17cb7f77342d765ba5024c40c0; native 1024x1024, 1:1; one request, first output; steps=40, guidance=3.0, seed=2582507270187158929; mps for text encoder and denoising; cpu for native VAE decode; bfloat16 for text encoder and transformer; float32 for VAE, generated Nov 2, 2026, 7:04 PM. Source
Protocol gradually-image-comparison-v1. Prompt SHA 256: 2e012c3d14ee5dd02203a94f394ca80a5038cdebd6032b25a8b8a221f2255419
SenseNova U1.5 Lite image SHA 256: df5c50fc54f03b972796f424999bd5991ff758a6b7abc00c212864aca3b124bc
Chroma1-HD image SHA 256: d0ccb26f1a27cc3142c3c669b3e05ed7bb568b58706a237e7e34b3bbb68a9cc5
The numbers at a glance
SenseNova U1.5 Lite
not in the arena
Chroma1-HD
not in the arena
SenseNova U1.5 Lite
Rank 14 of 108
Chroma1-HD
Rank 94 of 108
SenseNova U1.5 Lite
no API price
Chroma1-HD
no API price
Ranks among the 21 image models with an arena rating and the 108 with a GraduallyBench run; prices compared with 28 models with an API price.
SenseNova U1.5 Lite
- Subscription
- n/a No plan in Germany names exactly this model.
- Via API
- No API price known. Access: API, Web app, local.
- License
- Open weights.
Chroma1-HD
- Subscription
- n/a No plan in Germany names exactly this model.
- Via API
- No API price known. Access: local.
- License
- Open weights.
Subscriptions only count plans that name exactly this model. Subscription prices include VAT, API prices are in dollars.
GraduallyBench from Nov 3, 2026 · Subscription database from Sep 29, 2026 (38 days old)
All data side by side
The better value is bold. For ratings only when the gap is clear.
| Kennzahl | SenseNova U1.5 Lite | Chroma1-HD |
|---|---|---|
| Ratings | ||
| Arena ratingfrom votes by users, higher is better | n/a | n/a |
| Arena rank | n/a | n/a |
| Arena votesmore votes, a safer rating | n/a | n/a |
| GraduallyBenchour own rating, 0 to 100 | 73 | 23 |
| GraduallyBench by task | ||
| Typography and layoutLegible text, hierarchy, and composition | 93 | 36 |
| Product photographyMaterials, reflections, and fine details | 84 | 16 |
| Character and detailAnatomy, clothing, and spatial consistency | 35 | 17 |
| InfographicNumbers, labels, and visual organization | 80 | 22 |
| API price | ||
| Per imagelower is better | n/a | n/a |
| 100 images | n/a | n/a |
| Profile | ||
| Released | n/a | n/a |
| Weightsopen means: runs locally | open | open |
| Access | API, Web app, local | local |
| Parameters | not published | 8.9 billion |
| Largest output | 4K | not consistently published |
| Plans in Germanynaming exactly this model | n/a | n/a |
API prices for one image at 1,024 × 1,024 pixels according to Artificial Analysis.
Sources: SenseNova U1.5 Lite model source (verified Nov 2, 2026) · Chroma1-HD model source (verified Nov 1, 2026)
Arena ratings by category
From blind votes on LMArena. “Ahead” means the lead is clear.
No shared arena score in the stored snapshot
The stored benchmark database contains no measurement for SenseNova U1.5 Lite and Chroma1-HD from the same arena and category. Older model versions are not used as substitutes.
More comparisons
Compare SenseNova U1.5 Lite and Chroma1-HD with the models people compare them with most often.