Claude Fable 5.2 is now available. Anthropic's Mythos-class model for demanding reasoning and long-horizon agents, with a 1M-token context and 128K output. 65% off official pricing: $2.80 in / $14.00 out per 1M tokens (official $8 / $40), cache reads $0.07.Pricing & specs
Minimax

MiniMax M3

Save 70%Official

minimax/minimax-m3

MiniMax M3 — 1M context, reads images, thinking can be turned off.

Get an API keyView code sample
Live model playground

Try MiniMax M3 here

Tune the request, see the current price ceiling, and inspect the real response without leaving this model page.

streamchat
1Configure input
105 / 20,000
Add images

Sign in first to upload images to this vision model.

Maximum estimated cost—
2Live resultExample
MiniMax M3

Streaming responses

Pros:

  1. Lower perceived latency: Users see the beginning of the answer immediately, which can make interactions feel faster.
  2. Long responses feel more manageable: Text appears progressively, making it easier to follow or stop early.

Cons:

  1. More complex implementation: Clients must handle partial output, reconnection, and interrupted generations.
  2. Harder to validate and process: The full response is not available at once, complicating structured-output parsing, error handling, and some workflows.

Non-streaming responses

Pros:

  1. Simpler integration: The entire response is returned in one payload, making parsing, validation, and downstream processing easier.
  2. More predictable completion: The request finishes atomically, which is useful for batch jobs and applications that require the complete result before proceeding.

Cons:

  1. Higher perceived latency: Users may wait without seeing any output until generation is complete.
  2. Poor experience for long answers: A long response can appear frozen and provides no incremental feedback.
Example output · generated Oct 24, 2026 · took 4.7s · cost $0.000085
Context window
1M
Input modalities
TextImage
Output modalities
Text
Capabilities
Vision · Reasoning · Tool calling · Coding · Prompt cache
Upstream release
2026-06-01
RouteMux name
minimax/minimax-m3
Official ID
MiniMax-M3

What is MiniMax M3

MiniMax M3 is an M-series language model MiniMax released on June 1, 2026. Its release notes position it for agentic reasoning, tool use, coding, multimodal chat input and long-context work. MiniMax says its own MSA (MiniMax Sparse Attention) is what takes the context window to 1M tokens, cutting per-token compute at 1M context to about 1/20 of the previous generation, and that mixed-modality training from the first step gives it native image understanding.

On RouteMux it takes text and image input, supports tool calling and prompt caching, and answers on OpenAI Chat Completions, OpenAI Responses and Anthropic Messages. Pricing is tiered: requests with up to 512K input tokens bill at the standard rate, and anything larger bills entirely at the long-context rate. A priority tier is available by setting service_tier to priority. Thinking tokens count toward max_tokens, so a low limit can end the reply with max_tokens before any visible text; leave room for both the reasoning and the answer.

MiniMax released M3.1 Flash on October 19, 2026, with the same 1M context. The concrete reasons to stay on M3 are interface differences. Thinking can be switched off: MiniMax's docs say M3 skips thinking and answers directly when thinking is set to disabled, while M3.1 Flash always thinks. service_tier priority puts requests ahead in the queue, and M3.1 Flash does not accept that field. Behavior also stays pinned to the June release. In the other direction, MiniMax documents reasoning_effort depth levels for M3.1 Flash only; on M3 thinking is simply on or off.

On RouteMux the standard tier, up to 512K input tokens, is $0.09 per 1M input tokens and $0.36 per 1M output tokens, against MiniMax's $0.3 and $1.2.

MiniMax M3 API pricing

Prices verified 2026-10-20

InputOfficial $0.3 / M tokens
$0.09/ M tokensSave 70%
OutputOfficial $1.2 / M tokens
$0.36/ M tokensSave 70%

Fast $0.135 / $0.54 /M

Cache read $0.018/M · Cache write —/M

TierInputCache readCache writeOutput
Standard ≤512K$0.09$0.3−70%$0.018$0.06−70%—$0.36$1.2−70%
Standard >512K$0.18$0.6−70%$0.036$0.12−70%—$0.72$2.4−70%
Priority ≤512K$0.135$0.45−70%$0.027$0.09−70%—$0.54$1.8−70%
Priority >512K$0.27$0.9−70%$0.054$0.18−70%—$1.08$3.6−70%

Estimate cost

Estimated cost$0.00126

Key capabilities

1M-token context
Built on MSA sparse attention; requests over 512K input tokens bill at the long-context rate.
Thinking can be turned off
With thinking disabled it answers directly. Depth levels exist only on M3.1 Flash.
Priority tier
Set service_tier to priority to jump the queue; the rate is in the pricing section. M3.1 Flash does not support it.
Images and tools
Text and image input, tool calling and prompt caching, on all three protocols.

Limited time10% extra Credits on your first top-up — up to 200 bonus Credits

Applies to your first top-up within 72 hours of signing up. Minimum $20, once per account.

Sign up and claimNo subscription. Pay for what you use.

Interface and protocols

OpenAI Chat CompletionsStreamingOpenAI ResponsesStreamingAnthropic MessagesStreaming

Use any of these IDs to call this model via the API.

RouteMux nameminimax/minimax-m3
Official IDMiniMax-M3

Replace the ROUTEMUX_KEY placeholder with your API key. Create one →

from openai import OpenAI

client = OpenAI(
    base_url="https://api.routemux.com/v1",
    api_key="<ROUTEMUX_KEY>",
)

completion = client.chat.completions.create(
    model="minimax/minimax-m3",
    messages=[{"role": "user", "content": "What is the meaning of life?"}],
)
print(completion.choices[0].message.content)
base_url: https://api.routemux.com/v1

Performance (last 30 days)

Related models

MiniMax M3 FAQ

›What is MiniMax M3?

MiniMax M3 — 1M context, reads images, thinking can be turned off.

›How large is the context window of MiniMax M3?

MiniMax M3 supports a context window of up to 1M tokens.

›How much does the MiniMax M3 API cost?

Through RouteMux, MiniMax M3 costs $0.09 for input and $0.36 for output (USD per 1M tokens), below the official list price.

›How do I call MiniMax M3 via API?

MiniMax M3 is OpenAI-compatible: point your base URL at RouteMux and set the model field to minimax/minimax-m3 — no code changes needed.

›Which input and output modalities does MiniMax M3 support?

MiniMax M3 accepts text, image as input and produces text as output.

›When was MiniMax M3 released?

MiniMax M3 was released on June 1, 2026.

Sources

Reviewed by the RouteMux team on 2026-10-24

View as Markdown