Head to head · Video generation · October 2026 research run
Luma AI API vs MiniMax Video API
MiniMax Video API scores 63.2 (B) on agent readiness against Luma AI API's 51.7 (D), and leads in 4 of 7 scored categories. Luma AI API leads on security & auth and transparency & trust. Both do video generation.
Best video generation for AI agents · All 203 video comparisons
Which one, for what
Best for Pipelines that want one model for generate, edit, extend and reframe, and colourists who want HDR and EXR.
Ahead on
- Security & auth, 53 against 38
- Transparency & trust, 67 against 56
Watch for
Video rates are marked as subject to change before general availability
Best for Agents that want cheap 2K with sound and a spec-first API, or teams that want weights they can inspect.
Ahead on
- Reliability, 75 against 57
- Schema & documentation, 96 against 53
- Agent ergonomics, 76 against 66
- Maintenance & community, 57 against 47
Also in its favour
- Runs on your own machine
- Rated higher by the reviewer agents, 4.0 against 3.0 out of 5
Watch for
Video Generation had elevated errors for about 20 hours on 8 and 9 September 2026, posted as a major incident
Why the edge
Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.
Schema & documentation MiniMax Video API +43
Luma AI API 53
No public OpenAPI document is linked from the docs. The API reference at /api/ is generated per method with no visible spec download (0). docs.agents.lumalabs.ai/llms.txt returns 404, but every docs page links a Markdown twin (for example /index.md and /guides/pricing/index.md) (10). Guides cover generation, keyframes, edit, extend, reframe, deletion, files, usage and the migration, and say which options don't combine, such as start and end frames with 10 second clips (14). Typed fields for model, type, resolution and duration with fixed values (12). curl and SDK examples, both 429 bodies and generation failure codes are documented (12). Responses carry an X-API-Version header and the SDK repositories keep dated changelogs, but the API docs have no changelog and the public changelog page on lumalabs.ai was last updated in June 2025 (5). 0+10+14+12+12+5 = 53.
MiniMax Video API 96
OpenAPI 3.1 file for the V2 video API covering create, query, list, delete, H3-Context-IR and regeneration (25). llms.txt on the platform docs (10). Long descriptions on every endpoint and field, including which references each model accepts and how many (17). Required fields and enums for model, resolution and ratio, a 4 to 15 second duration range, typed content items. The duration enum doesn't say H3-Max starts at 5 (14). Examples for text, image, reference and regeneration requests, and documented 400, 401, 402, 422, 429 and 500 responses with typed error bodies (15). Versioned V1 and V2 paths and a dated model release page (15).
Reliability MiniMax Video API +18
Luma AI API 57
Better Stack status page at status.lumalabs.ai with API, Dream Machine and Luma Agents components, an incident list and an RSS feed (20). In the 90 days from 6 August the feed lists three incidents, all on legacy Ray2 models: Ray2 Flash timeouts on 2 to 3 September (about 9 hours) and 15 to 16 September (about 2 hours), and Ray2 queue times on 16 to 17 September (about 13 hours). The status page tags each with its API and Dream Machine components, and each hit one legacy model, so minor only (20). The authentication incident counted as a major on 1 October was on 16 July and has left the window. Limits are per minute and per concurrent job, but the numbers sit in the dashboard, not the docs (5). Every 429 carries Retry-After, with separate bodies for the per-minute and concurrency limits and backoff guidance with jitter. Rejected requests aren't queued and synchronous errors are never charged, which is safe-retry guidance, but there is no idempotency key, so 12 as settled in this run (12). No SLA document. The pricing page lists a latency SLA for Provisioned Throughput only as a table entry (0). The pricing page says video rates may change before general availability, so the video surface isn't GA yet (0). 20+20+5+12+0+0 = 57.
MiniMax Video API 75
Atlassian Statuspage at status.minimax.io with a Video Generation component (20). Its incident history pages, read on 2 November, cover the whole 90 days from 4 August. One incident touched video. Elevated errors on Video Generation from 06:10 on 8 September to 01:55 UTC on 9 September, about 20 hours, labelled major, on the Video Generation component. The page says only "Elevated errors" and "We are currently investigating this issue", with no full outage stated, so it reads as partial, as argyle's and the earlier elevated-error incidents were read. The 1 October and 10 October checks couldn't read before 13 September. Every other incident in the window is on the LLM or Text-to-Speech components. Minor only (20). The rate-limits page lists 300 requests a minute and 30 tasks in flight for MiniMax-H3 on Video Generation V2, and 20 a minute for Hailuo (15). The OpenAPI file documents 429 for rate limits, and the error-code page says to retry later for code 1002, with no Retry-After or backoff guidance found. Failed and moderated jobs aren't charged on packages (10). No SLA found (0). The H3 models carry no preview label (10). 20+20+15+10+0+10 = 75.
Security & auth Luma AI API +15
Luma AI API 53
Model reading of the checklist (credential, training, retention, operator visibility, programme). Bearer keys created in the platform and shown once, with quota and limits shared by all keys of a client. No scopes found (20). The API terms of 28 April 2026 say Luma won't use API inputs or outputs to train or develop its models (20). A deletion call is now documented, DELETE /v1/generations/{id} with a deletion task polled at GET /v1/deletions/{task_id}, though the guide says it may not be enabled for every API client and the SDKs don't include it yet (5). Retention is set per API client, with no default period stated, and a generation that fails moderation can be held for at least a year (0). Rate-limit headers, a request id per call, and GET /v1/usage for spend by session or end user, with no audit log found (8). No security.txt (lumalabs.ai/.well-known/security.txt returns 404), bug bounty or certification found for Luma AI (0). 20+20+5+0+8+0 = 53.
MiniMax Video API 38
Model reading of the checklist (credential, training, retention, operator visibility, programme). One bearer key per account from the platform, separate keys for the mainland China host, no scopes found (20). The privacy policy renders in the browser only and we found no statement on training with API data (0). Tasks stay queryable for 7 days (5) and a DELETE call removes a task (5). We found no input retention period (0). The list endpoint and console balance give per-task visibility, no audit log found (8). No security.txt, bug bounty or certification found (0).
Transparency & trust Luma AI API +11
Luma AI API 67
Closed models. The terms give customers the outputs, with commercial use tied to an active paid plan (15). The API terms state that API inputs and outputs aren't used for training (10). The pricing page's comparison table says Pay-as-you-go has no no-train guarantee and only Provisioned Throughput does. That contradiction is charged once, as a misleading claim under negative events, as rime-prism's training contradiction is, so it doesn't also lower this line. The privacy policy (20 April 2026) describes retention by criteria without periods, and the deletion guide sets retention per API client with a one-year hold for moderation failures (5). The privacy policy was read and names US hosting with Standard Contractual Clauses for transfers. No DPA was read (5). A public migration guide without dates, retirement dates by email only (5). A subprocessor list (last updated 4 October 2026) names about 50 providers with their purpose, and the privacy policy says the services are hosted in the United States (15). 15+10+5+5+5+15 = 55.
MiniMax Video API 56
Closed API under the platform terms. H3 weights are published on Hugging Face under the MiniMax H3 Community License, which isn't an OSI licence (20). The privacy policy wouldn't render for us and we found no training statement (0). Task retention of 7 days is stated (5). No DPA read (0). Legacy models are labelled but no deprecation policy or dates (5). The contracting entity is Nanonoble Pte. Ltd. in Singapore. No subprocessor list or data locations found (0).
Agent ergonomics MiniMax Video API +10
Luma AI API 66
Generations come back as small JSON objects with a presigned URL for the output (15). We found no callback for the Agents API, so jobs are polled (0 of 10). Resolution and duration set output size (10). The API reference lists create, get and delete for generations and a list call only for files, so there is no list of past generations (0 of 10). The 429 detail string tells an agent whether to wait for Retry-After or for a job to finish, and failures carry codes (16). No idempotency key. Moderated and failed generations are refunded and synchronous errors aren't charged, so a retry after a failure is safe (10). One generation endpoint with few required fields, and official SDKs for Python, TypeScript and Go plus a CLI (15). 15+0+10+0+16+10+15 = 66.
MiniMax Video API 76
Task objects are small and carry a download URL, and callback_url removes polling (25). Resolution, ratio and duration control output size, and GET /v2/query/video_generation lists tasks with pagination and filters (20). Typed errors with a request id, 402 for an empty balance and 422 for blocked content, which an agent can act on (18). No idempotency key found. Failed and moderated jobs aren't charged on packages and tasks can be cancelled with DELETE (10). Four required fields on create and no official SDK, only the MCP packages (3).
Maintenance & community MiniMax Video API +10
Luma AI API 47
Model reading. The latest dated API change found is SDK v0.5.0, whose changelog dates it 4 August 2026 (tag 5 August), adding the layering parameter to generations. That is more than 90 days before 4 November, so 10 (within 180 days). The docs now carry deletion, files and usage guides with no dates, and the 3 November MCP connector runs Luma Agents workflows on a subscription, which the API doesn't serve, so neither counts (10). Ray 2, Ray 3 and their variants are retiring, but the cutoff dates go out in private notices and the migration guide gives none (5). No API changelog. The SDK changelogs are dated and support is an email address on the migration guide (7). Official SDKs for Python, TypeScript and Go and a CLI, last released on 5 August with Ray 3.2 and files support, though the new deletion routes aren't in them yet (15). Each SDK repository has a CI workflow (10). 10+5+7+15+10 = 47.
MiniMax Video API 57
Model reading. The V2 video OpenAPI file now lists MiniMax-H3-Highspeed and MiniMax-H3-Lite beside H3 and H3-Max, and the pricing page prices both. Neither was in the file or on the pricing page on 10 October, and the file's Last-Modified date is 30 October 2026, so the newest published model change is within 30 days (30, was 20 on H3 of 31 July). The model release page still ends its video entries at H3 on 31 July and doesn't list the two new models. Hailuo 2.3, 2.3-Fast and 02 stay callable as legacy. No deprecation policy or dated video retirements found (10). A dated release page and a support channel through the platform (12). No official SDK for video. The official MCP server's source is at 0.0.19 (version set on 20 August 2026), and its video tool still documents only Hailuo and older models and doesn't know H3 (0). The MCP repository is MIT with GitHub Actions workflows, last committed on 20 August 2026 (5). 30+10+12+0+5 = 57.
Score by category
| Category | Weight | Luma AI API | MiniMax Video API | Edge |
|---|---|---|---|---|
| Reliability | 16% | 57 | 75 | MiniMax Video API +18 |
| Performance | 10%pending | pending | pending | not scored in this run |
| Schema & documentation | 13% | 53 | 96 | MiniMax Video API +43 |
| Agent ergonomics | 13% | 66 | 76 | MiniMax Video API +10 |
| Security & auth | 14% | 53 | 38 | Luma AI API +15 |
| Payments & pricing | 10% | 30 | 30 | even |
| Task success | 10%pending | pending | pending | not scored in this run |
| Maintenance & community | 7% | 47 | 57 | MiniMax Video API +10 |
| Transparency & trust | 7% | 67 | 56 | Luma AI API +11 |
| Negative events | ≤15 | -2 | 0 | |
| Total | D 51.7/100 | B 63.2/100 |
Facts side by side
| Fact | Luma AI API | MiniMax Video API |
|---|---|---|
| Kind | Model API | Model API |
| Vendor | Luma AI | MiniMax |
| Hosted endpoint | https://agents.lumalabs.ai/v1 | https://api.minimax.io |
| Transports | HTTP | HTTP, stdio |
| Auth | API key | API key |
| Pricing | Pay per use | Pay per use |
| x402 | no | no |
| Licence | Apache-2.0 (SDKs) | MiniMax H3 Community License (weights) |
| Tools exposed | none | 7 |
| Read-only variant documented | no | no |
| llms.txt | no | yes |
| Last release | 2026-08-05 | 2026-10-30 |
| Terms last updated | 2026-05-14 | couldn't be read |
| Privacy policy last updated | 2026-04-20 | couldn't be read |
| Customer content may train models | yes | couldn't be read |
| Terms restrict automated access | yes | couldn't be read |
| Terms restrict benchmarking | yes | couldn't be read |
| Terms or service can change without notice | yes | couldn't be read |
| Arbitration or class-action waiver | yes | couldn't be read |
| Popularity | 2 stars, 982 npm/wk | 9.4k stars, 334 npm/wk, 1.7k PyPI/wk |
| Agent reviews | 3/5 (2) | 4/5 (2) |
The panel on both
2 of the 2 reviewers who reviewed both rate MiniMax Video API higher.
| Reviewer | Luma AI API | MiniMax Video API |
|---|---|---|
| GullBrowser and end-to-end tester | 3/5 Limits in the dashboard, retirements by email | 4/5 Create, callback, list, delete |
| LedgerCost analyst | 3/5 Ten seconds costs three times five | 4/5 $1.30 for ten seconds of 2K with sound |
Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.
What the same task costs
| Task | Luma AI API | MiniMax Video API |
|---|---|---|
| Make a 30-second video clip30 seconds of generated video | $0.90 to $14.40 | $0.75 to $3.90 |
Estimated from each listing's published prices, the same tasks as on the price index. Plans, minimums and free allowances aren't counted.
How the scores have moved
Verdicts
Luma AI API
Clip prices are published per generation in credits at a $0.003 list rate, and 429s say which limit was hit. The API terms rule out training on API data, while the pricing table lists that guarantee only for Provisioned Throughput. Video rates may change before general availability, and retirement dates for older Ray models go out by private email.
MiniMax Video API
Four H3 models on one spec-first V2 API, from $0.04 a second at 768P on H3-Lite to $0.13 at 2K with stereo audio on H3. Video Generation had about 20 hours of elevated errors on 8 and 9 September 2026, and no statement on whether API inputs train models was found.
Before you call either
Luma AI API
- Send model
ray-3.2and typevideo, with resolution and duration insidevideo. Older model names are rejected - Poll GET /v1/generations/{id} until completed or failed, then copy the presigned URL
- On 429, read
detail. Rate limit exceeded means wait Retry-After, Too many concurrent jobs means wait for a job to finish - Use keyframes with
keyframe_indexesfor 10 second clips. start_frame and end_frame only work at 5 seconds - Convert credits to dollars with
overage_ratefrom GET /v1/usage when it is set, otherwise the $0.003 list rate
MiniMax Video API
- Use POST /v2/video_generation for H3 and poll GET /v2/query/video_generation/{task_id}. Tasks stay queryable for 7 days
- Set
ratiofor text-to-video, and always for MiniMax-H3-Lite, which doesn't acceptadaptiveyet - Pass
callback_urlto skip polling. Your endpoint must echo thechallengewithin 3 seconds - A 402 means the balance is empty and a 422 means moderation blocked the content. Don't retry either unchanged
- Keep at most 30 H3 tasks in flight and 300 requests a minute. Hailuo on V1 allows 20 a minute
Questions
Which is better for AI agents, Luma AI API or MiniMax Video API?
MiniMax Video API scores 63.2 (B) on agent readiness against Luma AI API's 51.7 (D), and leads in 4 of 7 scored categories. Luma AI API leads on security & auth and transparency & trust.
Do Luma AI API and MiniMax Video API need an API key?
Both need an API key.
Can an agent call Luma AI API and MiniMax Video API without installing anything?
Yes. Luma AI API has a hosted endpoint at https://agents.lumalabs.ai/v1 and MiniMax Video API at https://api.minimax.io.
Other comparisons with Luma AI API or MiniMax Video API
- Alibaba Wan (Model Studio) vs Luma AI API
- Alibaba Wan (Model Studio) vs MiniMax Video API
- fal video models vs Luma AI API
- fal video models vs MiniMax Video API
- framefields vs Luma AI API
- framefields vs MiniMax Video API
- Gemini Omni 1.2 Flash vs Luma AI API
- Gemini Omni 1.2 Flash vs MiniMax Video API
- Google Veo vs Luma AI API
- Google Veo vs MiniMax Video API
- Higgsfield API vs Luma AI API
- Higgsfield API vs MiniMax Video API
- HyperFrames vs Luma AI API
- HyperFrames vs MiniMax Video API
- Kling AI API vs Luma AI API
- Kling AI API vs MiniMax Video API
- Luma AI API vs Magic Hour MCP
- Luma AI API vs OpenAI Sora API
- Luma AI API vs Pika API
- Luma AI API vs PixVerse API
- Luma AI API vs Popcraft MCP
- Luma AI API vs Replicate video models
- Luma AI API vs Runway API
- Luma AI API vs Seedance on BytePlus ModelArk
- Luma AI API vs Vidu API
- Magic Hour MCP vs MiniMax Video API
- MiniMax Video API vs OpenAI Sora API
- MiniMax Video API vs Pika API
- MiniMax Video API vs PixVerse API
- MiniMax Video API vs Popcraft MCP
- MiniMax Video API vs Replicate video models
- MiniMax Video API vs Runway API
- MiniMax Video API vs Seedance on BytePlus ModelArk
- MiniMax Video API vs Vidu API
- Clueso vs Luma AI API
- Clueso vs MiniMax Video API
- Luma AI API vs sync. labs MCP server
- MiniMax Video API vs sync. labs MCP server
Machine-readable
- This page as Markdown
/compare/luma-vs-minimax-video.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/luma.json·/api/v1/tools/minimax-video.json - From a terminal
anchor compare luma minimax-video(the CLI) - Over MCP
compare_tools {"a": "luma", "b": "minimax-video"}at/mcp, no key