whatmodel.sh / video / LTX-2 19B Distilled
LTX-2 19B Distilled
↗ share
+ contribute
LTX-2 19B Distilled is an open-weights video model from Lightricks, released January 2026. Its published size is 19B parameters. Licence: ltx-2-community.
Model information sources
· Weights licence ↗ · Parameter count ↗
Lightricks' LTX-2 paper (maker). Card lists the checkpoint as 'ltx-2-19b-distilled | The distilled version of the full model, 8 steps, CFG=1' and ltx.io/llm-info says 'LTX-2: The original release. Approximately 19 billion parameters.' 19B = video+audio transformer streams; Gemma text encoder and VAEs separate.
$ how it compares
The video models ranked either side of LTX-2 19B Distilled, out of 139 with a score, then more from Lightricks.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 81 |  Wan 2.2 TI2V 5B | 47±2 | api price unknown +local | ||
| 82 |  Kandinsky 5.0 T2V Lite | 47±2 | local | ||
| 83 | LTX-2 19B Distilled | 47±8 | api $0.020/sec +local | ||
| 84 |  Cosmos-Predict2.5-14B | 47±3 | comparison only | ||
| 85 |  CogVideoX-5B-I2V | 46±4 | local | ||
| more from Lightricks | |||||
| 53 | LTX-2.5 | 53±6 | api $0.12/sec +local | ||
| 70 | LTX-2 | 49±2 | local | ||
| 112 | LTX-2.3 | 36±3 | api $0.040/sec +local | ||
$ what it's good at
text to video▆▄▃▄▆▅█▃▂▂▂▄ #38 of 68 * · 75% of its score · 1 test
Beats Wan 2.2 TI2V 5B here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Sci-VBench testmini (expert human rating) | 42.4% ✎ | ▅·▃█▃·▅▃▅▃▃▃ beats 2/14 | 75% |
$ where to run it
Through an API
api · Replicate · $0.020/sec · replicate.com ↗ · price sourceUSD per second of output video on Replicate's official model lightricks/ltx-2-distilled ('LTX-2 Distilled is a 19 billion parameter model'; billing table: $0.02 per second of output video, one flat tier; no hardware-time charge). Product: Replicate model API. API model: lightricks/ltx-2-distilled. Settings: text-to-video, default 1080p (the page says it generates 1080p by default and can render up to 4K; the flat rate applies to all sizes), audio generated. Price basis: published output-based list rate. Note: LTX's own API removed ltx-2-fast/ltx-2-pro (2026-08-16 changelog) and fal's fal-ai/ltx-2-19b/ endpoints carry an 'will be deprecated on August 15, 2026' notice; the Replicate official model shows no deprecation. Currency: USD. Not a Lightricks direct tariff. Replicate official model page, observed 2026-09-26.
Other ways to run it
version
- Run it as published · needs ~40.9 GB, Mac only · from github.com/drawthingsai
brew install draw-things-cli && draw-things-cli generate --model ltx_2_19b_distilled_q6p.ckpt --prompt "Ocean waves at sunset" --frames 49 --seed 42 --video-format h264 --output clip.mp4
Official weights and documentation: huggingface.co/Lightricks ↗
$ how the score was worked out · Sci-VBench: Evaluating Knowle… 75%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science Domains | 1 | 75% |
| prior · pull toward the middle | 25% | |
| weighted average | = 47.0 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json