MiniMax Pricing 2026: API Costs, Token Plans & Credits

Last verified: September 24, 2026

MiniMax pricing is divided into usage-based API billing, monthly Token Plans, prepaid Credits, Audio Subscriptions, and Video Packages. For MiniMax-M3 Standard requests with no more than 512K input tokens, the currently published billed rates are $0.30 per 1 million input tokens, $1.20 per 1 million output tokens, and $0.06 per 1 million prompt-cache read tokens.

Other published rates include $60 per 1 million characters for speech-2.8-turbo, $0.38 per hour for Speech Recognition (ASR), $0.08 per output second for MiniMax H3 at 768P, $0.13 per output second for H3 at 2K, $0.05 and $0.08 per output second for the new MiniMax H3 Max at 480P and 768P, a retained rate of $0.15 for up to five minutes of Music 3.0 for eligible existing paying API users, $0.0035 per image for image-01, and $0.01 per request for both API-vlm and the web_search server tool. Effective August 20, 2026, MiniMax closed paid Music Generation and Lyrics Generation APIs to new users and discontinued the free Music API variants. Hailuo 2.3, Hailuo 2.3 Fast, and Hailuo 02 pricing is retained below as legacy V1 pricing.

Important Music API Availability Update

Effective August 20, 2026: MiniMax’s official service-adjustment notice says paid Music Generation and Lyrics Generation APIs are no longer available to new users. Existing paying users may continue using the current API services. The free APIs music-3.0-free, music-2.6-free, and music-cover-free are discontinued.

For new access, MiniMax directs users to MiniMax Audio or the open-weight MiniMax Music 3 checkpoint. A retained price or model enum is not evidence that a new account can activate the API.

Important API-vlm Pricing Update

MiniMax reduced the published API-vlm price from $0.06 to $0.01 per request, effective July 22, 2026. MiniMax states that the API endpoints and model capabilities did not change, so the pricing update does not require a code change.

When API-vlm is called through Token Plan, usage is deducted from the included Token Plan quota according to the pay-as-you-go price. Therefore, the lower per-call price also reduces the amount deducted for each eligible Token Plan call. Check the official MiniMax Pay as You Go pricing page before deployment or purchase.

MiniMax Pricing at a Glance

Billing option	Published starting price	Best fit	Credential or balance
Pay as You Go	M3 from $0.30/M input tokens and $1.20/M output tokens; H3 Max video from $0.05/output second; Speech Recognition $0.38/hour	Production APIs and usage-based workloads; the only billing route for H3 and H3 Max video	Standard API Key and Open Platform wallet
Token Plan	$20/month	Individual interactive development and coding workflows	Subscription Key and included quota
Purchased Credits	$5 for 5,000 Credits	Eligible usage without a plan seat or overflow beyond included quota	Subscription Key and Credits balance
Audio Subscription	$5/month	Recurring speech-generation workloads	Audio points and plan limits
Video Package	$1,000 for one month	Listed Hailuo legacy video workflows; verify current model coverage	Prepaid video points

Practical choice: use Pay as You Go when an application needs measured production billing. Token Plan is designed for individual, interactive use and has rolling and weekly quota windows. Audio and Video Packages are most relevant after measuring a sustained workload in that modality.

API Key, Subscription Key, Wallet, and Credits

Item	Used for	Deduction source	Important rule
Pay-as-you-go API Key	Usage-based API calls	Open Platform wallet	Separate from the Subscription Key
Subscription Key	Token Plan and purchased Credits	Included quota first, then eligible Credits	Can exist before a plan seat or Credits are assigned, but has no usable paid resources in that state
Open Platform wallet	Pay-as-you-go balance	Actual measured API usage	Each Team has its own wallet
Credits	Eligible Subscription Key usage	1,000 Credits equal $1 for deduction	Separate from the Open Platform wallet

Do not switch keys blindly after an authorization or insufficient-balance error. First determine whether the request should consume an Open Platform wallet, an included Token Plan quota, or purchased Credits. See the independent MiniMax Token Plan guide for setup details.

MiniMax Pay-as-You-Go API Pricing

The following prices are in US dollars. “M tokens” means 1 million billable tokens. Input, output, and prompt-cache usage have separate published rates.

MiniMax-M3 Standard and Priority Pricing

Service tier	Input-token tier	Input / 1M tokens	Output / 1M tokens	Cache read / 1M tokens
Standard	≤512K input tokens	$0.30	$1.20	$0.06
Standard	>512K input tokens	$0.60	$2.40	$0.12
Priority	≤512K input tokens	$0.45	$1.80	$0.09
Priority	>512K input tokens	$0.90	$3.60	$0.18

• MiniMax labels the displayed M3 billed rates as “Permanent 50% off” and shows higher crossed-out reference prices.
• The applicable M3 row depends on whether the request contains no more than 512K or more than 512K input tokens.
• Priority is enabled with service_tier: "priority" and is priced at 1.5 times Standard.
• The official table describes Priority as providing priority admission, faster responses, and improved request reliability. It does not publish a guaranteed latency figure.
• The pricing table publishes M3 cache-read prices but does not list a separate M3 prompt-cache write price.

Read the MiniMax M3 guide before designing a long-context workload. Context-window capacity should not be treated as a target input size; longer requests increase token consumption and can use the higher pricing tier.

MiniMax-M2.7 Pricing

Model	Input / 1M	Output / 1M	Cache read / 1M	Cache write / 1M
MiniMax-M2.7	$0.30	$1.20	$0.06	$0.375
MiniMax-M2.7-highspeed	$0.60	$2.40	$0.06	$0.375

The high-speed entry has higher input and output prices. Both hosted entries are documented with a 204,800-token context window. See the MiniMax M2.7 guide for model and compatibility details.

Legacy M-Series Pricing

MiniMax places M2.5, M2.1, and M2 in the legacy section of its model catalog. Published legacy prices do not mean that a legacy model is the preferred choice for a new application.

Legacy model	Input / 1M	Output / 1M	Cache read / 1M	Cache write / 1M
MiniMax-M2.5	$0.30	$1.20	$0.03	$0.375
MiniMax-M2.5-highspeed	$0.60	$2.40	$0.03	$0.375
MiniMax-M2.1	$0.30	$1.20	$0.03	$0.375
MiniMax-M2.1-highspeed	$0.60	$2.40	$0.03	$0.375
MiniMax-M2	$0.30	$1.20	$0.03	$0.375

Speech, Speech Recognition, Video, Music, Image, and Tool Pricing

Speech 2.8 and Voice Pricing

API or feature	Model scope	Published price	Billing note
Text to Speech	speech-2.8-turbo	$60 / 1M characters	Published for supported TTS usage
Text to Speech	speech-2.8-hd	$100 / 1M characters	Published for supported TTS usage
Rapid Voice Cloning	Listed speech models	$1.50 / voice	The cloning fee is charged on the first synthesis using the cloned voice
Voice Design	Listed speech models	$3 / voice	The design fee is charged on the first synthesis using the designed voice

Synchronous Text to Speech supports up to 10,000 characters per request. The asynchronous long-text workflow supports up to 1 million characters per request. The limits describe request capacity; actual character usage determines pay-as-you-go TTS cost.

The older speech-2.6-turbo and speech-02-turbo entries remain listed at $60 per 1 million characters. Their HD variants remain listed at $100 per 1 million characters. See the Speech 2.8 guide for endpoint-specific details.

Speech Recognition (ASR) Pricing

MiniMax now lists an Automatic Speech Recognition API in the Audio section of its Pay-as-you-go page. It turns recorded or streamed audio into text and is billed by audio duration rather than by characters.

API	Endpoint	What it includes	Published price
Speech Recognition	Speech-to-Text	Audio-to-text transcription with streaming, speaker diarization, and subtitle export in SRT or VTT format	$0.38 / hour

• MiniMax’s rate-limit page lists the model as asr-1.0 and says Speech to Text is available to paid accounts only, not on the free tier.
• The published limits are 2 concurrent connections and 30,000 TPM, where TPM counts audio seconds, so it caps how much audio can be transcribed per minute.
• At the list price, a 90-minute recording costs about $0.57 and 10 hours of audio costs $3.80.

MiniMax H3 Video Pricing

MiniMax-H3 is one of two current video models in MiniMax’s main catalog, alongside MiniMax-H3-Max. It uses Video Generation V2 and currently requires Pay as You Go. The official Token Plan documentation excludes H3.

H3 billing item	Published rate	Billing note
768P generation	$0.08 / output second	4–15 second output
2K generation	$0.13 / output second	4–15 second output
Audio reference	Free	When supplied as an H3 reference input
Image references	First 5 free; then $0.04 / image	Per generation request
Reference video	Per input second at the selected output-resolution rate	Input duration is billable
768P-to-2K regeneration	$0.05 / output second	Reference-input charges can also apply
H3 Context-IR	$0.90 / 1M input tokens; $3.60 / 1M output tokens	Separate token-based context processing

A ten-second H3 output costs $0.80 at 768P or $1.30 at 2K before any billable reference inputs. See the independent MiniMax H3 guide, the official Pay-as-you-go table, and the official H3 V2 API reference.

MiniMax H3 Max Video Pricing

MiniMax-H3-Max is the speed-optimized video model that fal.ai post-trained on H3 and that MiniMax describes as jointly released with fal.ai. It outputs 480P or 768P clips of 5–15 seconds; 2K is not available. It is Pay-as-you-go only: MiniMax’s video guide and V2 API reference both say H3 Max requires the Pay-as-you-go API, so a Token Plan Subscription Key or Credits balance is not the route for it.

H3 Max billing item	Published rate	Billing note
480P generation	$0.05 / output second	5–15 second output
768P generation	$0.08 / output second	5–15 second output
Audio input	Free	Reference audio
Image input	First 2 free; then $0.074 / image	First-frame, last-frame, and reference images
Video input at 480P output	$0.0553 / input second	Billed by input-video duration
Video input at 768P output	$0.143 / input second	Billed by input-video duration

Worked example: ten 10-second text-to-video clips cost $5.00 on H3 Max at 480P (10 × 10 × $0.05) and $8.00 at 768P, the same as H3 at 768P. Input fees can reverse the comparison: a 15-second 768P request with four reference images and a 6-second reference video costs about $2.21 on H3 Max ($1.20 output + $0.148 for two extra images + $0.858 of input video), compared with $1.68 on H3, which includes five free images and bills that input video at $0.08 per second.

Third-party hosts such as fal set their own H3 Max prices and free allowances, which are not MiniMax prices. The independent MiniMax H3 Max guide compares the routes, the speed claims, and more cost scenarios.

Legacy Hailuo Video API Pricing

Model	Resolution	Duration	Published price per video
MiniMax-Hailuo-2.3-Fast	768P	6 seconds	$0.19
MiniMax-Hailuo-2.3-Fast	768P	10 seconds	$0.32
MiniMax-Hailuo-2.3-Fast	1080P	6 seconds	$0.33
MiniMax-Hailuo-2.3	768P	6 seconds	$0.28
MiniMax-Hailuo-2.3	768P	10 seconds	$0.56
MiniMax-Hailuo-2.3	1080P	6 seconds	$0.49

Legacy model notice: MiniMax now places Hailuo 2.3, Hailuo 2.3 Fast, and Hailuo 02 under Legacy Models. Their V1 identifiers and prices may remain documented for existing integrations, but new video workflows should evaluate MiniMax-H3 or MiniMax-H3-Max on V2 first. Hailuo 2.3 supports documented text-to-video and image-to-video workflows; Hailuo 2.3 Fast is image-to-video only.

Legacy MiniMax-Hailuo-02 remains listed at $0.28 for 768P/6s, $0.56 for 768P/10s, $0.49 for 1080P/6s, $0.10 for 512P/6s, and $0.15 for 512P/10s.

MiniMax Music Pricing and Availability

Model or feature	Published price or status	Availability from August 20, 2026
Music-3.0	$0.15 / up to 5 minutes	Closed to new users; existing paying users may continue. The official pricing table labels the service “Discontinued” while retaining the rate.
Music-2.6	$0.15 / up to 5 minutes	Closed to new users; existing paying users may continue. The official pricing table labels the service “Discontinued” while retaining the rate.
Lyrics Generation	$0.01 / song	Closed to new users; existing paying users may continue. The official pricing table labels the service “Discontinued.”
music-3.0-free, music-2.6-free, and music-cover-free	Discontinued	The free Music APIs are no longer available under MiniMax’s August 20 service notice.

MiniMax’s current pricing table labels Music 2.5+, Music 2.5, and Music 2.0 as discontinued while preserving historical rates of $0.15 for Music 2.5+ and Music 2.5 and $0.03 for Music 2.0, each for music of up to five minutes. These retained figures are not evidence of new-account access or a recommended integration.

Image, API-vlm, and Web Search Pricing

Area	Model or tool	Published price	Cost driver
Image	image-01	$0.0035 / image	Per generated image; a request can ask for 1–9 images
Multimodal MCP/API tool	API-vlm	$0.01 / request	Per call
Server tool	web_search	$0.01 / request	Per server-side search call

If image-01 returns image URLs, those URLs expire after 24 hours. Download any output that must be retained. Read the Image-01 guide for supported output dimensions and request options.

MiniMax Token Plan Pricing

Token Plan is a monthly subscription accessed through a Subscription Key. The current plan uses usage-based deduction and a unified included quota across supported API Platform resources. The console usage bar is the source of truth for remaining quota. Included quota is controlled by a five-hour rolling window and a weekly window.

H3 and H3 Max coverage notice: the official Token Plan documentation currently excludes MiniMax H3, and MiniMax’s video guide and V2 API reference say both H3 and H3 Max require Pay as You Go. Do not interpret a generic video allowance in a plan comparison as H3 or H3 Max access.

Plan	Monthly price	Official best-fit description	Approximate agent guidance	Listed video generation
Plus	$20	Personal projects and daily workflows	Approximately 3–4 agents	Not included in the current plan comparison
Max	$50	Daily professional coding and multimodal work	Approximately 4–5 agents	3 generations per day
Ultra	$120	Heavy daily use and extended agent workflows	Approximately 6–7 agents	5 generations per day

• The agent figures are approximate usage guidance, not guaranteed simultaneous production capacity.
• Supported model usage shares one included quota rather than a separately published monthly token allowance for every model.
• Unused included quota does not carry into the next billing cycle.
• A plan can be upgraded during the subscription period by paying the price difference; the upgrade takes effect immediately.
• The Token Plan is designed for individual, interactive developer use. MiniMax recommends Pay as You Go for production.
• MiniMax may apply dynamic rate limiting during peak traffic.
• Check the account console before purchase because model access and quotas can change.

Review the official Token Plan price table and the official Token Plan FAQ for current deduction and quota rules.

Purchased Credits

Price	Credits received	Conversion	Validity
$5	5,000 Credits	1,000 Credits = $1; eligible usage is deducted at the corresponding pay-as-you-go price	365 days from each purchase date
$25	25,000 Credits
$100	100,000 Credits

Credits can be purchased and used without an assigned Token Plan seat. They still require a Subscription Key and have the same resource coverage as Token Plan. If both included quota and Credits can cover a request, included quota is used first and Credits cover eligible overflow.

Token Plan for Teams

• The Team Owner can purchase seats and assign each seat to one member.
• A seat can be reassigned during the billing cycle, but the current usage state follows the seat and does not reset.
• A Team Owner can purchase a shared Credits pool.
• Members without an assigned seat can use shared Credits if access is enabled.
• Pay as You Go remains separate and uses the wallet belonging to the Team where the standard API Key was created.

The official Token Plan for Teams guide explains seat assignment, shared Credits, and Team wallet behavior.

MiniMax Audio Subscription Pricing

Audio Subscriptions publish monthly audio points, voice slots, rate limits, and subscription totals. MiniMax does not publish a universal one-to-one conversion between audio points and pay-as-you-go billable characters, so the table should not be converted into an unsupported effective per-character price.

Plan	Monthly	3 months	Yearly	Points/month	Voice slots	RPM
Starter	$5	$13.50	$48	100,000	10	10
Standard	$30	$81	$288	300,000	100	50
Pro	$99	$267	$950	1,100,000	250	200
Scale	$249	$672	$2,390	3,300,000	500	500
Business	$999	$2,697	$9,590	20,000,000	800	800

The official table labels three-month subscriptions as 10% off and yearly subscriptions as 20% off. Custom pricing is available through MiniMax. Confirm renewal and checkout terms on the official Audio Subscription page.

MiniMax Video Package Pricing

Package	Discount label	Price	Validity	Video points	RPM
Standard	Save 5%	$1,000	1 month	3,760	20
Pro	Save 10%	$2,500	1 month	9,920	30
Scale	Save 15%	$4,500	1 month	18,900	40
Business	Save 20%	$6,000	1 month	26,780	50

Video-Point Deductions

Model	Configuration	Points per generation
MiniMax-Hailuo-2.3-Fast	768P / 6s	0.7
MiniMax-Hailuo-2.3-Fast	768P / 10s	1.1
MiniMax-Hailuo-2.3-Fast	1080P / 6s	1.3
MiniMax-Hailuo-2.3 or MiniMax-Hailuo-02	768P / 6s	1
MiniMax-Hailuo-2.3 or MiniMax-Hailuo-02	768P / 10s	2
MiniMax-Hailuo-2.3 or MiniMax-Hailuo-02	1080P / 6s	2
MiniMax-Hailuo-02	512P / 6s	0.3
MiniMax-Hailuo-02	512P / 10s	0.5

Under the Video Package rules, failed generations and videos that trigger security review do not deduct points. Unused package points do not carry over and reset to zero when the package expires. The point table above covers the listed Hailuo models and does not establish H3 eligibility. Confirm current model coverage and terms on the official Video Packages page before making a high-value purchase.

MiniMax API Cost Examples

For an uncached language-model request without paid tools, use:

estimated_cost =
(input_tokens / 1,000,000 × input_rate)
+ (output_tokens / 1,000,000 × output_rate)

Example 1: M3 Standard Below 512K

A request uses 100,000 input tokens and produces 20,000 output tokens, without a cache discount or paid tool:

Input: 100,000 / 1,000,000 × $0.30 = $0.030
Output: 20,000 / 1,000,000 × $1.20 = $0.024
Estimated total: $0.054

The same usage at the published Priority rates would cost $0.081.

Example 2: API-vlm

Five hundred API-vlm calls at $0.01 per call cost:

500 × $0.01 = $5

Example 3: Speech 2.8 Turbo

Generating 250,000 billable characters with speech-2.8-turbo costs:

250,000 / 1,000,000 × $60 = $15

Example 4: MiniMax H3 at 2K

Ten ten-second H3 outputs at 2K, before any billable reference inputs, cost:

10 videos × 10 seconds × $0.13 = $13.00

Example 5: MiniMax H3 Max at 480P

The same ten ten-second outputs on H3 Max at 480P, before any billable inputs, cost:

10 videos × 10 seconds × $0.05 = $5.00

Example 6: Speech Recognition

Transcribing a 90-minute recording with speaker diarization and an SRT export at the ASR list price costs:

1.5 hours × $0.38 = $0.57

These examples cover only the published units shown. Actual account charges can differ because of cache usage, output length, additional calls, tools, multiple generated images, taxes, currency conversion, or account-specific terms.

MiniMax Pricing FAQ

How much does MiniMax-M3 cost?

For Standard requests containing no more than 512K input tokens, the published billed rates are $0.30 per 1 million input tokens, $1.20 per 1 million output tokens, and $0.06 per 1 million prompt-cache read tokens. Higher prices apply above 512K input tokens or when Priority is selected.

How much does API-vlm cost?

The published API-vlm price is $0.01 per request, effective July 22, 2026. MiniMax states that the price change did not alter the API endpoints or model capabilities.

Is the MiniMax API free?

Most production API usage is billed or consumes eligible subscription resources. As of August 20, 2026, MiniMax says the free Music APIs music-3.0-free, music-2.6-free, and music-cover-free are discontinued, Speech to Text is limited to paid accounts, and H3 and H3 Max video require Pay as You Go. Product trials, promotional credits, and open-weight downloads are separate from a generally available free API and can carry their own limits, costs, and licenses.

What is the difference between Pay as You Go and Token Plan?

Pay as You Go uses a standard API Key and deducts actual usage from an Open Platform wallet. Token Plan uses a Subscription Key and provides an included monthly subscription quota controlled by rolling and weekly windows. MiniMax recommends Pay as You Go for production use.

Does Token Plan include MiniMax H3 or H3 Max?

No. The current official Token Plan documentation excludes H3, and MiniMax says both H3 and H3 Max must be used through the Pay-as-you-go API. Generic video allowances shown for a plan should not be interpreted as H3 or H3 Max entitlement. Check the official documentation and account console before purchase because coverage can change.

How much does MiniMax H3 Max cost?

H3 Max output costs $0.05 per second at 480P and $0.08 per second at 768P. Audio inputs are free, the first two images are free and each extra image costs $0.074, and input video costs $0.0553 per second at 480P or $0.143 per second at 768P. A 10-second 480P clip therefore costs $0.50 before inputs.

How much does MiniMax speech recognition cost?

The Speech Recognition (Speech-to-Text) API is listed at $0.38 per hour of audio and includes streaming, speaker diarization, and SRT/VTT subtitle export. MiniMax says it is available to paid accounts only.

Can Credits be used without a Token Plan?

Yes. Official MiniMax documentation states that Credits can be purchased and used without an assigned Token Plan seat. They still use a Subscription Key and cover resources within Token Plan coverage.

Do unused Token Plan quotas roll over?

No. Unused included Token Plan quota does not carry over into the next billing cycle. Purchased Credits have a separate validity of 365 days from each purchase date.

Does MiniMax charge for prompt caching?

MiniMax publishes prompt-cache read rates for M3, M2.7, and the listed legacy language models. M2.7 and the listed legacy entries also have published prompt-cache write rates. The central pricing table does not list a separate M3 cache-write rate.

Do the listed prices include tax?

This independent guide cannot determine the tax, billing currency, payment fee, exchange rate, enterprise discount, or contractual price that applies to an account. The official checkout or contract is the final source.

---

How This Pricing Guide Is Verified

This page was last checked against MiniMax’s live pricing, video, and rate-limit documentation on September 24, 2026, when the H3 Max and Speech Recognition entries were added. It prioritizes MiniMax’s operational pricing tables, Token Plan FAQ, model catalog, and account-facing documentation over third-party calculators or undated summaries. Prices, discounts, quota windows, RPM limits, model access, and package entitlements can change. Confirm any purchase or production estimate through official MiniMax documentation and the account console.

Continue with the MiniMax API guide, MiniMax model directory, or application-building guide.