Grok Imagine Video 1.5 Lite: $0.14 a Second at 1080p and a Spot Above Veo 3.1

Grok Imagine Video 1.5 Lite is xAI's cheaper video tier, announced October 8, 2026, at $0.02 per second for 480p, $0.03 for 720p and $0.14 for 1080p. Independent benchmarker Artificial Analysis placed it 17th on its text-to-video leaderboard the same evening, two spots above Google's Veo 3.1, at about a third of Veo's per-second price. The official numbers come from the Imagine account on X and the xAI models page, which lists the Lite model at $0.02 per second; the ranking comes from Artificial Analysis.

Quick facts

Note on sourcing: the price and rank figures above are as reported in the xAI Imagine and Artificial Analysis posts on X as of October 8. We could not open the original posts directly during writing, so treat the exact values as reported and check the live pricing page before budgeting.

What changed

xAI, now operating under the SpaceXAI banner in much of the press, already sells Grok Imagine Video 1.5. We covered that release in June in our Grok Imagine Video 1.5 launch breakdown: a 720p image-to-video model with native audio that topped a blind image-to-video arena at the time. The Lite model is a different product decision. Rather than push quality, it pushes cost and latency, and it adds 1080p, which the June release lacked.

Two price points matter:

• Against the full model. Artificial Analysis lists the full Grok Imagine Video 1.5 at $0.25 per second at 1080p. At $0.14, Lite is 44 percent cheaper, in exchange for a lower rank (Artificial Analysis puts the full model six places higher).
• Against Google. Veo 3.1 is listed at about $0.40 per second at 1080p, so Lite costs about 35 percent as much while ranking slightly higher on this specific leaderboard.

At the low end, $0.02 per second means a 10 second 480p draft costs about 20 cents, which is the real story for anyone iterating on prompts before spending on a final render.

How to read the leaderboard result

Artificial Analysis runs a text-to-video arena where people compare anonymous clips, then publishes a quality ranking alongside price and speed. Rank 17 is not first place, and it should not be read as "better than Veo" across the board. Three caveats:

1. It is the silent board. The reported ranking is on the AA-Video-T2V v2.0 list for silent output. Audio, lip sync and sound design are a separate comparison, and the benchmarker itself notes weakness in lip sync.
2. Rank gaps near the middle are small. Two places on a crowded board usually means a modest preference difference, not a generational gap.
3. Veo 3.1 is not the newest Google video model. Google has shown newer work; see our notes on Gemini Omni video and how it compares in Runway Aleph 2 versus Gemini Omni. Beating an older price-heavy model on cost is the main win here.

The more durable claim is the Pareto one: Artificial Analysis says no other model on its silent list is both faster and higher quality than Lite. That means if you plot quality against generation time, Lite sits on the frontier. For a production pipeline that renders many clips, speed (60.5 seconds median for 10 seconds of 1080p) matters about as much as the score.

Where it is available

The launch coverage lists three routes: xAI's own Imagine API, Vercel AI Gateway, and fal. Gateways matter because they let teams swap video models behind a single integration and compare cost directly. xAI already put its image model there, as we noted in the Grok Imagine Image 2 on Vercel AI Gateway post. Check model names and aliases on the xAI models page before you hard-code anything.

Who should use it

• Prototyping and storyboards. Sub-5-cent-per-second pricing at 720p makes it easy to generate many options.
• Product and social clips with on-screen text. Artificial Analysis says text rendering and lighting are strengths, which is where many video models stumble.
• Programmatic video at scale. The speed profile helps queues finish faster.

Who should wait or test first:

• Anything with people talking. The reported lip sync and anatomy weakness is a real limit for ads and explainers with faces.
• Anything needing audio. The ranking says nothing about it.
• Anyone needing a stable API contract. A new tier on a fast-moving platform can change names and prices.

Context: xAI's shipping pace

This launch is one of several xAI releases in recent months, from the Grok 4.6 launch and evals to Grok 4.7. Video is where xAI has leaned on price: the June model undercut rivals by a wide margin, and Lite continues that. The strategy is consistent, even if each release needs independent confirmation of the quality claims.

How to try it, step by step

1. Pick a route. If you already use xAI, call the Imagine API directly. If you run several providers, the gateway routes (Vercel AI Gateway or fal) let you change a model string instead of rewriting code.
2. Start at 480p. Write the prompt, generate a 5 second clip at $0.02 per second, and judge motion and composition. Resolution does not change what the model understood from your prompt, so a cheap draft tells you most of what a final render will.
3. Promote only the keepers. Re-render the best prompts at 1080p. Because the same prompt can produce a different clip on each run, save the seed or parameters if the API returns them, and expect some variation.
4. Keep text short. Even strong text rendering degrades with long strings. Ask for a single word or short phrase on a sign or label and check spelling frame by frame.
5. Avoid close-up dialogue. If a shot needs a face to speak convincingly, plan to use a separate lip-sync or audio step, or choose a model that scores better on that dimension.
6. Log cost per usable clip. Divide total spend, including failed attempts, by the number of clips you actually keep. That figure, not the headline per-second price, is what decides whether Lite is cheaper than a pricier model that needs fewer retries.

Why cheap video models matter now

Video generation has long been the most expensive modality to serve, which is why per-second pricing, rather than per-token pricing, dominates the market. Each price cut changes who can experiment: a marketing team that would not spend $6 on a 15 second concept clip may happily generate fifty drafts at under $1 each. It also changes the competitive picture for the labs. Google, OpenAI, Runway and others have all pushed quality, while xAI has repeatedly competed on price and speed, first with the June release and now with a Lite tier. For readers following the wider shift in how AI companies price their products, our coverage of Claude Sonnet 5.5 cache-read price cuts shows the same pressure on text models.

A leaderboard position and a low price are also easy to quote and hard to verify in the abstract. Artificial Analysis publishes methodology for its arenas, but the outcomes depend on the prompt mix and on which voters show up. The honest summary is that Lite looks competitive for its price, and that your own prompts are the only test that counts.

A quick cost sketch

Using the listed rates: a 15 second clip at 1080p costs about $2.10 on Lite versus $6.00 at Veo 3.1 per Artificial Analysis pricing, and $3.75 at the full Grok Imagine Video 1.5 rate. A 100-clip batch of 10-second 720p drafts at $0.03 per second costs $30. These are straight multiplications of the rates in the table, and they exclude retries, which in practice raise the effective cost because video models often need several attempts per usable clip.

What to verify before you commit

1. Open the xAI docs and confirm the model identifier, resolution tiers and the 1 to 15 second range.
2. Run ten of your own prompts through Lite and your current model and compare faces, hands and any on-screen text.
3. Measure end-to-end latency in your region rather than relying on a median.
4. Check content and licensing terms for commercial use.

Bottom line

Grok Imagine Video 1.5 Lite is a pricing and speed move: a 1080p tier at $0.14 per second that Artificial Analysis ranks just above Veo 3.1 on a silent text-to-video board. It is attractive for drafts, text-heavy clips and bulk generation, and not yet proven for talking people. Verify with your own prompts before switching.

Related reading

• Grok Imagine Video 1.5 launch, June 2026
• Grok Imagine Image 2 on Vercel AI Gateway
• Grok 4.6 launch and evals
• Google Gemini Omni video model
• Runway Aleph 2 vs Gemini Omni

Primary: xAI models and pricing docs; the Imagine account (@imagine) and Artificial Analysis (@artificialanlys) posts of October 8, 2026.

Details are accurate as of October 9, 2026. Prices and rankings change often; check the live docs.