MiniMax ships open-weight video model H3, repricing video from per-clip to per-second
MiniMax launched MiniMax-H3, an open-weight omni-modal video model that replaced Hailuo as its video product, moving video billing from $0.19-$0.56 per clip to $0.08-$0.13 per second.
Hailuo 2.3 video billed per clip at $0.19-$0.56 depending on resolution (768P-1080P) and duration (6-10s); no video-regeneration or context-IR SKUs.
MiniMax-H3 video bills per second: $0.08/sec at 768P, $0.13/sec at 2K. New MiniMax-H3-Regeneration SKU (768P to 2K upscale) at $0.05/sec, and MiniMax-H3-Context-IR per-token pricing ($0.90/M in, $3.60/M out). Token Plan subscriptions explicitly exclude MiniMax-H3 from included quota coverage.
MiniMax’s international pricing docs now lead the video section with MiniMax-H3, described as an open-weight, general-purpose, omni-modal generation model. It replaces the Hailuo-branded video line entirely on the live site and documentation — the Hailuo name no longer appears anywhere in MiniMax’s product navigation as of this capture, and the consumer showcase app previously branded Hailuo AI now appears as MiniMax Hub.
The pricing mechanic changed along with the rebrand: Hailuo 2.3 billed per finished clip (a flat $0.19-$0.56 depending on resolution and duration), while MiniMax-H3 bills per second of generated output ($0.08/sec at 768P, $0.13/sec at 2K) — a genuine unit-of-metering change, not just a relabel. MiniMax also added two new video SKUs: a regeneration tier that upscales a previously produced 768P clip to 2K at $0.05/sec, and an H3-Context-IR task priced per token ($0.90/M input, $3.60/M output) rather than per second.
One entitlement detail is notable for Token Plan subscribers: the plan’s monthly quota page explicitly states MiniMax-H3 is “not currently supported” under the included quota, alongside voice design and rapid voice cloning — so H3 usage always draws from unrestricted Credits or pay-as-you-go billing, even on a paid subscription.
Live USD international card: flagship MiniMax-M3 and MiniMax-M2.7 at $0.30/M in / $1.20/M out (cache read $0.06), M3 >512k and M2.7-highspeed at $0.60 / $2.40; M1/M2/M2.5 moved to 'Legacy Models.' Music 3.0 launches at $0.15/track (free tier at RPM 3). New per-call surfaces: MCP API-vlm cut $0.06→$0.01/call (effective 2026-07-22) and server-tool web_search at $0.01/request, both deducting from Token Plan quota then Credits. Token Plan held at $20/$50/$120 with a '2× tokens' revamp (up to ~12.5B tokens/month).