Kling 4.0 vs Grok Imagine Video: The Cheapest Model Has the Most Votes

The most heavily measured model in this comparison is not the newest one, and the newest one has not been measured at all. Grok Imagine Video is a family of three entries on the Artificial Analysis image-to-video board: the January 2026 original at Elo 1,072 with 14,371 votes behind it, Grok Imagine Video 1.5 at 1,098 with 5,267 votes, and the Lite variant on the text-to-video board. Kling 4.0, released on 15 October 2026, has no row on any board, no published per-second rate, and a spec sheet that beats all three Grok entries on duration, resolution and reference count. The gap between evidence and ceiling is the whole matchup.

That shape is more common in this market than the leaderboard format admits. A model with a long vote history is a known quantity, not necessarily the best one. A model eight days old with a 30-second pass is a promise with a demo reel. The question is which kind of uncertainty your project can absorb.

Three Grok models, one Kling

Grok Imagine Video is not one product, and the version numbers do not rank the way you would expect. Read on the image-to-video board with audio on, as captured on 10 October 2026:

• Grok Imagine Video original — 13th at Elo 1,072 ±7 over 14,371 samples, January 2026, $4.20 per minute.

• Grok Imagine Video 1.5 — 9th at Elo 1,098 ±8 over 5,267 samples, May 2026, $8.40 per minute.

• Grok Imagine Video 1.5 Lite — 16th on the text-to-video board at 993 ±10 over 5,715 samples, released 26 August 2026, $8.40 per minute.

• Kling 4.0 — no row on any board as of 23 October 2026.

The original is the cheapest and the best-measured. The 1.5 refresh scores 26 Elo higher and costs twice as much per minute, on a third of the votes. The Lite variant sits on a different board entirely, so comparing it to the other two is already comparing tasks. This is a family that has been iterated on a consumer platform's cadence, and the price ladder reflects that: you can buy the cheap, well-evidenced baseline, the dearer refinement, or the cheaper-per-second Lite tier, and each one asks you to trade evidence against score.

Kling 4.0 offers none of that internal choice. There is one model, one unreleased price, and one duration ceiling. That simplicity is a feature if you want the best available take and a liability if you want a procurement decision you can defend with a number.

What the vote counts buy you

Sample count is usually treated as a technicality. On this board it is the difference between a claim and a guess. The January Grok checkpoint's ±7 interval is the same width as Gemini Omni Flash's, which has 12,327 votes behind it; 14,371 comparisons is enough for the board to state a position with some confidence. Grok 1.5's ±8 is wider because it has fewer votes, not because it is a worse model. Kling 4.0's interval does not exist yet.

That has a practical consequence for anyone choosing on quality. If you pick Grok Imagine Video original, you know roughly what you are getting: a mid-table model with a strong evidence base and the lowest rate in this comparison. If you pick Kling 4.0, you are accepting the vendor's claims on a model that independent evaluators have not yet scored. Neither choice is irrational. They are different bets.

The spec ceiling

On paper, Kling 4.0 is the more capable machine by a clear margin.

• Duration — Kling 4.0 generates 3 to 30 seconds in one pass, with two-minute extension listed as coming soon. Grok Imagine Video 1.5 generates 1 to 15 seconds, with autoregressive extension from the last frame; the original is capped at 8.7 seconds for edit and extend work.

• Resolution — Kling 4.0 lists 720p, 1080p and 4K, with 10-bit HDR at 1080p and 4K marked coming soon. Grok 1.5 outputs 480p, 720p and 1080p natively. The original tops out lower.

• References — Kling 4.0 takes up to 15 assets across images, videos, subjects and voices, with caps of 10 images, 5 videos, 7 subjects and up to 3 video-based subjects. Grok 1.5 takes up to 14 reference images, up to 3 voice references, and up to 4 mid-video keyframes plus first and last frames.

• Keyframes — Kling 4.0 takes up to 10 keyframe images. Grok 1.5 takes first frame, last frame and up to 4 mid-video frames.

• Audio — both generate native audio with lip-synced speech; Kling claims two-channel stereo and nine languages plus dialects, Grok claims joint audio-video generation with up to 3 preset voice references.

• Prompt — Kling 4.0 accepts up to 8,000 tokens. Grok's documentation does not publish comparable prompt limits.

The two are closer on references than the headline numbers suggest: 15 assets against 14 images plus 3 voices is not a rout. The real gap is duration and resolution. Thirty seconds at 4K against fifteen at 1080p is a different class of deliverable, and it is the reason Kling 4.0 is worth considering at all despite the missing evidence.

Price: a published ladder against a blank

Grok Imagine Video's prices are published per second by resolution. The 1.5 model runs $0.08 per second at 480p, $0.14 at 720p and $0.25 at 1080p, with audio included and image input at $0.01 per image. The Lite variant runs $0.02 at 480p, $0.03 at 720p and $0.14 at 1080p. The legacy original is $0.05 at 480p and $0.07 at 720p. Artificial Analysis normalizes the 1.5 row to $8.40 per minute and the original to $4.20, which is consistent with the 720p rates.

Kling 4.0 has no vendor per-second rate as of 23 October 2026. The app runs on credits through Kling AI's plans; the API is in early access. Reseller listings quote per-second figures, but they are not Kuaishou's price and they vary widely enough that using one as a budget line would be guesswork.

For a real baseline, the previous Kling generation is priced and in production. On OrcaRouter, kling/kling-v3 and kling/kling-v3-omni are $0.084 per request, kling/kling-3-turbo is $0.112, and the older 2.x tiers run $0.042 to $0.28. Those are per-request prices rather than per-second rates, so they do not line up one-for-one with Grok's card, but they are the numbers a Kling buyer can actually use today.

Access and the edit surface

Grok Imagine Video is reachable through xAI's API and the consumer Grok product. The 1.5 model is asynchronous: submit, poll a request ID, download. The family has been served long enough that its request schema is documented and its failure modes are known. The original handles edit and extend work, capped at 720p and an 8.7-second source; the 1.5 model is generation-focused and explicitly does not support video editing or extension as documented features.

Kling 4.0 is live in the Kling AI app and website globally, with the global launch announcement dated 16 October 2026. The API is early access, with MCP and CLI support listed. Its editing surface is documented as targeted edits to expression, body movement, camera position, style and background across up to five input clips within a 30-second combined window — broader than Grok 1.5's documented editing support, narrower than a general-purpose video editor.

Neither model is an OrcaRouter route today, so nothing here is a claim that we host either. The Kling generation we do carry is Kling 3.0, which is the practical fallback if you are testing Kling 4.0 during early access. More broadly, the reason to hold a gateway in front of a video stack is that these models fail in different ways: one rate-limits, another times out on a long generation, a third returns a terminal status that is not a success. A single OpenAI-compatible endpoint with automatic failover and provider list prices passed through at 0% markup does not make the models better, but it stops a single bad endpoint from becoming a failed batch.

Who should pick which

• Pick Grok Imagine Video original if you want the lowest published rate and the strongest evidence base. At $4.20 per minute and 14,371 votes, it is the least surprising choice in the comparison, and it is cheap enough to use for drafts.

• Pick Grok Imagine Video 1.5 if you need native 1080p, 15-second takes, up to 14 reference images and 3 voice references, and a documented per-second price. It is the best-equipped Grok for a controlled character shot.

• Pick Kling 4.0 if you need a 30-second single pass, 4K, 10 keyframes or a 21:9 deliverable, and you are willing to pilot an unscored, unpriced early-access API.

• Pick Grok if procurement needs a number this week; pick Kling if the shot needs the longer take and you can wait for the rate card.

The honest conclusion is that these two models are not really competing for the same purchase yet. Grok Imagine Video is a mature, measured, cheap family sold on a public rate card. Kling 4.0 is a higher-ceiling model in the first week of general availability, with the evidence and the pricing still to come. The right move for most teams is to keep the measured model in production and pilot the new one on the shots that actually need 30 seconds — and to re-read both the board and the rate card in a month, when the blank column may finally have a number in it.