# MiniMax H3 vs H3 Max: frames, references and route limits Source: [MiniMax H3 vs H3 Max: frames, references and route limits]() Documentation index: [llms.txt]() Last updated: 2026-10-28 As reviewed on **28 October 2026**, ChatCut exposes **MiniMax H3** and **H3 Max** as different selectors. H3 exposes multimodal references and 768p/2K; Max exposes text/frame generation at 480p/768p and blocks references in the inspected build. The checked Max endpoint explicitly rejected last-only and audio-only requests. H3 generation success and Max in-limit reference completion remain unverified; neither model has a demonstrated quality advantage here. ## Exact models and dated evidence Tool aliases `minimax-h3` and `minimax-h3-max` map to the public IDs below. Native V2 documentation was accessed **27 October**; saved integration and endpoint evidence was reviewed **28 October 2026**. A validation response proves that particular rule, not a completed video or deployed production entitlement. | Model and public ID | Facts and evidence | | --- | --- | | **MiniMax H3**
`MiniMax-H3` **Availability:** `available_in_inspected_build`
**Restriction reason:** `unknown`
**Endpoint verification:** `validation_only` Pinned integration and saved dev checks; production unverified. Source links describe native specifications. **Reviewed:** 2026-10-28 | **Native documentation:** 4–15s; 768P/2K. First, last-only or both; OR up to 9 images, 3 videos, 3 audios, 12 total. Native audio-only references.
Native documentation; not an endpoint test; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) **ChatCut integration:** Frame and multimodal-reference modes exposed; do not mix them.
Pinned integration; production unverified; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) **Endpoint evidence:** Duration rejection enumerated 4–15s. Last-only/audio-only probes did not complete generation; ordinary H3 route success remains unverified.
Saved development evidence; not production certification; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) | | **MiniMax H3 Max**
`MiniMax-H3-Max` **Availability:** `available_in_inspected_build`
**Restriction reason:** `not_exposed`
**Endpoint verification:** `validation_only` Pinned integration and saved dev checks; production unverified. Source links describe native specifications. **Reviewed:** 2026-10-28 | **Native documentation:** 5–15s; 480P/768P. V2 documentation describes frames and multimodal references; no 2K tier.
Native documentation; not an endpoint test; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) **ChatCut integration:** Text/frame generation exposed; all Max references blocked. A last frame requires a first frame.
Pinned integration; production unverified; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) **Endpoint evidence:** Exact endpoint explicitly rejected last-only and audio-only input. Reference validators are not successful in-limit generation proof.
Saved development evidence; not production certification; 2026-10-28. [Native specification source 1 (2026-10-27)](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md) | ## Which input mode should I choose? Start with **text only**, **frames**, or **multimodal references**. Do not mix frames with reference arrays. MiniMax requires a nonempty prompt, with a documented 7,000-character ceiling. In frame mode, the source image determines the adaptive ratio. Native text mode lists 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16; reference mode also allows adaptive. These are specification/control facts, not measured output boundaries. [V2 task API](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md). For **H3**, the native reference budget is up to **9 images, 3 videos and 3 audios, 12 total**. Videos and audio are each 2–15 seconds, with at most 15 seconds total per type. Native limits include 30 MB per image, 50 MB per video and 15 MB per audio, plus a 64 MB request budget. ChatCut’s public-resource handling separately caps images at **5 MiB** and video/audio at **50 MiB**; the tighter applicable rule wins. See [generation references](https://pr-692-preview.chatcut.dev/docs/generation-references). Native H3 documentation permits a last frame alone and audio-only references. The saved H3 probes reached generic invalid-input errors, not a successful video. **Do not treat those modes as endpoint-proven.** On **Max**, last-only and audio-only were explicitly rejected by the configured endpoint. A last frame needs a first frame; reference audio needs an image or video on that route. The inspected ChatCut build blocks all Max references before submission, so native reference documentation is not a workaround. ## Does Max mean better quality, or include 2K? No matched quality or speed comparison was performed. MiniMax describes Max as a faster variant, which does not establish a completion-time promise. Choose its **480p** tier or **5–15-second** range when those controls fit your task. Choose H3 for its exposed **768p/2K** tiers or reference mode. The endpoint duration validators enumerated **4–15 seconds for H3** and **5–15 for Max**; accepted output duration and resolution were not measured by those rejections. Neither checked native V2 request offers a general seed or Seedance-style Draft-to-final switch. Do not sell a lower-resolution generation as a guaranteed reusable final render. ## Reproduce one small take 1 #### Choose a visible selector In Video Gen, select H3 or H3 Max actually shown in your build. For a simple Max test, use five seconds at 480p and 16:9. 2 #### Use one mode Start with text alone: “A red paper boat crosses a shallow puddle, static low camera, one continuous shot.” If using Max frames, attach a first frame before adding any last frame; do not attach references. 3 #### Review, then assemble Check the model, settings and current credit estimate before sending. A successful job returns a video asset to My Assets. Inspect motion, framing and sound before placing it on the timeline for trimming, captions and assembly. This is an actionable example, **not a newly executed H3/Max test**. Timeline edits remain reversible, but generated subjects are not separate editable layers. A rejected request calls for checking its mode, media and settings; it does not establish that the entire model is unsupported. ## How should I estimate credits? Use the current preflight quote and [Credits Policy](https://pr-692-preview.chatcut.dev/docs/credits-policy). Record the exact model, account cohort/catalog, resolution, ratio, output seconds, reference-video seconds and quote date; reconcile the settled charge in Credits history. Agent usage is separate. The current Max text/frame path quotes output seconds; its first/last pair falls within the native first-two-image allowance. A future reference-enabled route could also bill reference video and additional images. Native supplier rates and historical public-card estimates are not a current ChatCut invoice. [MiniMax pricing](https://platform.minimax.io/docs/pricing/overview.md). ## Frequently asked questions Can H3 Max use my image, video or audio references in ChatCut? The inspected build blocks all Max references, even though native V2 documentation describes them. Use H3 for the exposed reference mode; successful in-limit reference generation on the checked Max endpoint remains unverified. Why did a last frame alone fail on Max? The configured Max endpoint explicitly requires a first frame when a last frame is present. General MiniMax documentation does not override that exact route rejection. Does ordinary H3 prove that audio-only generation works? No. H3 documentation permits it, but the saved probes did not complete a video. H3 endpoint success remains unverified in this review. Can H3 Max generate 2K? The checked Max tiers are 480p and 768p. The separate H3 selector exposes 768p and 2K. Is Max the highest-quality H3 model? This review has no matched quality test. Choose by exposed input mode, resolution, duration and the current quote rather than an untested quality ranking. **Last verified: 28 October 2026.** Sources: [MiniMax V2 task API](https://platform.minimax.io/docs/api-reference/video-generation-v2-create.md), [video guide](https://platform.minimax.io/docs/guides/video-generation.md) and [pricing](https://platform.minimax.io/docs/pricing/overview.md), accessed 27 October. Pinned controls and saved endpoint validation were reviewed 28 October; production availability was not certified. Compare the [model matrix](https://pr-692-preview.chatcut.dev/docs/model-capability-matrix) and [video controls tutorial](https://pr-692-preview.chatcut.dev/docs/video-generation-models); use [workflow failures](https://pr-692-preview.chatcut.dev/docs/workflow-failures) for validation, timeout or missing-option diagnosis.