xAI's video-generation line has moved fast in 2026 — Grok Imagine Video 1.5 launched in June, known for pairing image-to-video animation with native, synchronized audio (music, sound effects, even lip-synced dialogue). This week's news isn't a new flagship, though — it's Grok Imagine Video 1.5 Lite, a distilled, cheaper, faster sibling that spent the days around October 8, 2026 climbing a major independent video-model leaderboard, and beating a Google model that costs several times more per second.
The headline comparison — Lite outranking Veo 3.1 on Artificial Analysis's video leaderboard — comes with real caveats worth understanding before you read it as 'better than Google.' Here's what's actually confirmed, and what the ranking does and doesn't tell you.
Quick summary: Grok Imagine Video 1.5 Lite, a distilled version of xAI's Grok Imagine Video 1.5, is priced in three resolution tiers — $0.02 per second at 480p, $0.03 per second at 720p (its native rendering resolution), and $0.14 per second at 1080p (upscaled from 720p) — all with audio included. It generates clips from 1 to 15 seconds across nine aspect ratios. On Artificial Analysis's AA-Video-T2V v2.0 leaderboard, Lite ranks #17 with an Elo score of 993 (±10) across 5,715 comparisons, two spots above Google's Veo 3.1 — while costing roughly 65% less than Veo 3.1's $0.40-per-second rate at matching 1080p-with-audio quality. Its full-size parent model, Grok Imagine Video 1.5, ranks higher still, at #11 with an Elo of 1045.
What Lite actually trades away
Grok Imagine Video 1.5 Lite is explicitly a distilled model — a smaller, faster version trained to approximate its larger parent's outputs rather than matching its full capability. It renders natively at 720p and reaches 1080p through upscaling rather than native high-resolution generation, and on the same leaderboard, it sits 52 Elo points behind the full Grok Imagine Video 1.5 (993 versus 1045). Both models support text-to-video and image-to-video generation with built-in audio, and both inherit the same underlying trick: text-to-video actually runs as text-to-image followed by image-to-video, generating a first frame from the prompt and then animating it (Sources: docs.x.ai, orcarouter.ai).
Pricing: three tiers, one clear budget pick
| Resolution | Price per Second (with audio) | Notes |
|---|---|---|
| 480p | $0.02 | Cheapest tier; native rendering |
| 720p | $0.03 | Lite's native resolution |
| 1080p | $0.14 | Upscaled from 720p, not natively rendered |
Even at its most expensive tier, Lite at $0.14 per second for 1080p-with-audio comes in well below Google's Veo 3.1, which charges roughly $0.40 per second for comparable 1080p output with audio — a gap of around 65%. That price difference is the real headline here, arguably more than the leaderboard rank itself: for teams generating large volumes of short-form video content where 'good enough' beats 'best possible,' Lite's pricing alone is a meaningful shift (Source: Alphasignal).
The leaderboard win, with its asterisk
Artificial Analysis's AA-Video-T2V v2.0 board ranks video models using blind pairwise preference voting — human raters see two outputs for the same prompt, without model names attached, and pick the one they prefer. On that measure, Grok Imagine Video 1.5 Lite sits at #17 with an Elo of 993 (±10), built from 5,715 head-to-head comparisons, placing it two spots above Veo 3.1. That's a real, independently-run result — but it reflects viewer preference on one specific prompt set and one snapshot of each model, not a technical fidelity or accuracy measurement, and leaderboard rankings shift as new prompt batches and model versions roll through. Treat 'beat Veo 3.1' as true for this specific leaderboard snapshot, not as a permanent or universal verdict (Source: orcarouter.ai).
If your use case is simple, well-defined clips at high volume — social content, product demo loops, ad variations — Lite's combination of low cost and a competitive leaderboard rank makes it worth testing first. If you need the most polished, highest-fidelity single output and cost matters less, the full Grok Imagine Video 1.5 (or a premium competitor) is still the safer default; Lite trails its own parent model by a real margin on the same leaderboard.
- Marketers and social teams producing high volumes of short clips where per-video cost compounds fast.
- Developers building video-generation features who want a cheaper fallback tier alongside a premium option.
- Anyone comparing xAI's video pricing against Google's Veo or OpenAI's Sora before committing to an API.
- Not a fit for anyone who needs native 4K or the single highest-fidelity output available — that's still the job for flagship-tier models, not a distilled Lite variant.
01How much does Grok Imagine Video 1.5 Lite cost?
Three tiers based on resolution: $0.02 per second at 480p, $0.03 per second at 720p (its native resolution), and $0.14 per second at 1080p (upscaled from 720p), all with audio included.
02Did Grok Imagine Video 1.5 Lite really beat Google's Veo 3.1?
On one specific independent leaderboard — Artificial Analysis's AA-Video-T2V v2.0, based on blind preference voting — yes, Lite ranked #17 versus Veo 3.1's lower spot, while costing roughly 65% less at comparable 1080p-with-audio quality. That reflects one benchmark's snapshot, not a universal verdict across every use case.
03What's the difference between Grok Imagine Video 1.5 and the Lite version?
Lite is a distilled, smaller, faster, and cheaper version of the full model. It renders natively at 720p (upscaling to reach 1080p) and scores about 52 Elo points lower than its full-size parent on the same leaderboard — a real quality gap in exchange for lower cost and higher speed.
04Does Grok Imagine Video 1.5 Lite include audio?
Yes — like the full Grok Imagine Video 1.5, Lite generates video with native synchronized audio, including music, sound effects, and lip-synced dialogue.
05What video lengths and formats does it support?
Clips from 1 to 15 seconds, across nine aspect ratios including widescreen 16:9 and vertical 9:16, supporting both text-to-video and image-to-video generation.
Video generation is moving as fast as text models did two years ago, and the cheapest option isn't always the right one for a given shot. LumiChats gives you access to 40+ premium AI models in one place — so while we're primarily built for comparing chat and reasoning models like Claude, GPT, Gemini, and Grok side by side, the same instinct applies here: test before you commit, because the leaderboard winner this week may not be the leaderboard winner for your specific prompt.
