MiniMax M3.1 Flash API: Release Date, Pricing, and How to Use It

MiniMax M3.1 Flash launched October 19. Get its exact API model ID, 1M context, $0.15/$0.60 pricing, three APIMaster routes, and a working integration example.

MiniMax M3.1MiniMax M3.1 FlashMiniMax-M3.1-FlashMiniMax APIcoding agentsAPIMaster

Published 2026-09-27 · Updated 2026-10-23

Quick Answer

MiniMax M3.1 Flash was officially released on October 19, 2026, and now has a documented public API. Its official model ID is MiniMax-M3.1-Flash. It supports a 1,000,000-token context window, text/image/video input, text output, and five reasoning-effort levels. The September MiniMax Code preview is historical access information; the earlier statement that M3.1 Flash has no public API is obsolete.

APIMaster lists 3 active public Flash routes at $0.15 input / $0.60 output per million tokens, checked October 23, 2026. All three match MiniMax's standard token price, so the current discount is 0%. Open the M3.1 Flash model card to compare current route terms. MiniMax's official cached-input price is $0.015 per million tokens; this is a vendor rate, not a verified APIMaster cache-billing quote. Live availability and prices on the marketplace card take precedence.

Prices below are USD per million input / output tokens. Route ranges count only public routes with status=1 at the check date.

Model ID	Original price	APIMaster current price	Active routes	Discount	Model card
`MiniMax-M3.1-Flash`	$0.15 / $0.60; cached input $0.015	$0.15 / $0.60 on all routes	3	0%; matches official standard pricing	View Flash routes
`minimax-m3`	Standard: ≤512K input $0.30 / $1.20; >512K $0.60 / $2.40	From $0.30 / $1.20; range $0.30–$0.60 / $1.20–$2.40	3	0% at the minimum versus the official ≤512K tier	Compare M3 routes
M3.1 without the Flash suffix; M3 Pro	No separate verified price in the sources checked	No public route found for `MiniMax-M3.1` or `MiniMax-M3-Pro`	0	—	—

MiniMax M3.1 Flash release date: what changed after the preview?

On September 27, MiniMax Code offered M3.1-Flash-Preview for everyday development. That product label did not establish a separately priced public API at the time. On October 19, MiniMax's official launch announcement introduced the released M3.1 Flash, while the current API documentation now names MiniMax-M3.1-Flash and publishes its pricing and request parameters.

MiniMax also confirmed that Space Bunny was an early version of M3.1 Flash. This resolves the anonymous preview's identity; it does not prove that every preview checkpoint is identical to the final release. For the preview history and old OpenCode ID migration, see the Space Bunny / M3.1 Flash guide.

The old September 28–October 7 check-in campaign has ended. It is not a current API entitlement, an APIMaster discount, or evidence of permanent free access. MiniMax Code subscriptions and pay-as-you-go API billing remain separate products.

MiniMax M3.1 Flash API model ID, context, and reasoning

The exact official API identifier is MiniMax-M3.1-Flash. Use that identifier for the Flash route shown on APIMaster; minimax-m3 selects the separate M3 model. Do not copy M3.1-Flash-Preview from an old Code selector into a new API integration.

Feature	Confirmed information	Practical implication
Context window	1,000,000 tokens	Budget input, conversation history, reasoning, and output within the endpoint's limits
Input / output	Text, images, video / text	Multimodal input does not mean image or video generation; audio input is not supported in the documented interface
Thinking	Always on	Disabling thinking or setting `reasoning_effort` to `none` is rejected by the official API
Effort levels	`low`, `medium`, `high`, `xhigh`, `max`; default `max`	Select effort explicitly to control latency and reasoning-token use
OpenAI-compatible controls	`reasoning_effort`, `reasoning_content`	Preserve reasoning and tool-call fields required by the endpoint in multi-turn agent sessions

Do not infer a universal maximum output allowance from the 1M context specification. Maximum output, timeouts, rate limits, multimodal forwarding, and cache billing can differ by gateway or route. Check the selected endpoint before running a large repository or long autonomous session.

MiniMax M3.1 Flash vs M3 and M3 Pro

Start by evaluating Flash for everyday coding and tool-driven agents. It has published API documentation and lower standard input/output prices than M3. For an existing M3 integration, compare completed-task quality, tool-call validity, retries, latency, and total cost before switching; lower token prices alone do not guarantee a better result on every task.

MiniMax reports 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82% on CyberGym for Flash. These are vendor-reported launch results, not APIMaster's independently reproduced tests. The launch post also reports nearly 3× faster inference and 70% lower overall agentic usage cost than its predecessor. The workload-cost claim is not an APIMaster token-price discount or a guaranteed speed on every route.

M3 Pro remains a separate name. This Flash release does not establish a launch, architecture, parameter count, or price for a distinct M3 Pro or unsuffixed M3.1 model. Earlier community claims about roughly 3T parameters and MSA 2.0 must not be presented as Flash specifications. Keep the M3 Pro status page and H3.1 status page separate from this released Flash API.

MiniMax M3.1 Flash API pricing on APIMaster

The October 23 marketplace check found 3 active public MiniMax-M3.1-Flash routes, all at $0.15 input / $0.60 output per million tokens. The input and output price ranges are therefore $0.15–$0.15 and $0.60–$0.60. That matches MiniMax's official standard price: 0% discount. Compare the live Flash routes for current capacity, supported features, and billing terms.

For a simple estimate, 1M uncached input tokens plus 100K billed output tokens costs $0.21 at these rates: $0.15 + 0.1 × $0.60. Count billable reasoning according to the endpoint's usage reporting, and check any additional charges separately. The official $0.015 cached-input rate should only enter an APIMaster budget after the selected route's cache pricing is confirmed.

The separate M3 card has 3 active routes, from $0.30 / $1.20 to $0.60 / $2.40. MiniMax's standard M3 pricing depends on whether input is ≤512K or >512K tokens. The cheapest route equals the official short-context rate; do not advertise a blanket 50% saving by comparing that route to the higher long-context tier without confirming eligibility. Flash's flat standard token prices avoid that particular vendor pricing split, while route limits still apply.

How to use the MiniMax M3.1 Flash API

Create an APIMaster account, create an API key, and open the Flash model card. Set your OpenAI-compatible base URL to https://apimaster.ai/v1 and use the exact Flash model ID:

curl "https://apimaster.ai/v1/chat/completions" \
  -H "Authorization: Bearer $APIMASTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-M3.1-Flash",
    "reasoning_effort": "medium",
    "max_tokens": 4096,
    "messages": [
      {"role": "user", "content": "Inspect this failing unit test and propose a minimal patch with a regression check."}
    ]
  }'

reasoning_effort is documented by MiniMax; verify that your selected gateway route forwards it. Start with medium for a small bug fix and compare the result with low or high. Thinking remains enabled at every supported effort level. For a direct MiniMax integration, use MiniMax credentials and https://api.minimax.io/v1; an APIMaster key does not authorize that separate service.

For agent workflows, preserve complete tool-call history and required reasoning fields, make side-effecting tools idempotent, and set a deliberate timeout. Validate one small request before moving a production coding workflow to the new model.

FAQ

Is MiniMax M3.1 Flash released, and when?

Yes. MiniMax officially released M3.1 Flash on October 19, 2026. The earlier September 27 availability was a MiniMax Code preview. The current release has a documented public API named MiniMax-M3.1-Flash.

Can I call MiniMax M3.1 Flash through APIMaster?

Yes. The October 23 check found 3 active public routes at $0.15 input / $0.60 output per million tokens. Use MiniMax-M3.1-Flash and check the live model card before integration.

What is the MiniMax M3.1 Flash price and discount?

MiniMax's standard rate is $0.15 input / $0.60 output / $0.015 cached input per million tokens. APIMaster's three active routes currently match the input/output rate, so the current discount is 0%. APIMaster cached-input billing was not verified in this marketplace check.

Can I disable MiniMax M3.1 Flash thinking?

No. The official API requires thinking and rejects reasoning_effort: "none" or thinking: {"type":"disabled"}. Supported effort levels are low, medium, high, xhigh, and max; the default is max.

Is M3.1 Flash the full M3.1 or the rumored M3 Pro?

The documented released model is M3.1 Flash. Its release does not confirm a separate unsuffixed M3.1 or M3 Pro API. Do not attach rumored 3T-parameter or MSA 2.0 specifications to Flash.

Is M3.1 Flash permanently free?

No permanent free API entitlement is established by the sources checked here. The September check-in promotion has ended, and any launch-free gateway offer has its own terms. APIMaster's current public Flash routes are paid at the rates above.

Sources and next step

MiniMax M3.1 Flash launch announcement — October 19 release, Space Bunny identity, vendor-reported benchmarks, and flat pricing.
MiniMax text-generation guide — model ID, 1M context, and supported models.
MiniMax OpenAI-compatible API guide — multimodal input, reasoning controls, and history handling.
MiniMax official pricing — Flash standard/cache prices and M3 context-dependent tiers.
APIMaster Flash card and M3 card — active-route pricing checked October 23, 2026; marketplace cards take precedence.

To try the released Flash API, register with APIMaster, create an API key, set the base URL to https://apimaster.ai/v1, and select MiniMax-M3.1-Flash. Compare real task results and route terms before scaling.

Back to Blog