MiniMax M3
MiniMax · released Jun 2, 2026 · MiniMaxAI/Minimax-M3
- Type
- Open weightsCustom licence
- Params
- 427B
- Context
- 1M
active per word not recorded by us · about 786K words of context · download allowed, licence restricts use
Our take
Written Aug 4, 2026MiniMax M3 is a large multimodal model whose weights can be downloaded. It accepts text, images and video, and scores well on the independent chat leaderboard we track for its price. It is a quality-per-dollar pick rather than a pure peak-quality one.
Choose this for high-volume chat or retrieval workloads where quality per dollar beats peak quality, or for multimodal image and video input at downloadable-model prices. Skip it if the restricted licence poses legal risk, or if you need frontier-level chat quality.
The case for it
- High measured chat score for the price: frontier models above it cost several times more.
- Video input at commodity pricing, with text and image input too.
The case against it
- Custom restricted licence; check terms before fine-tuning or redistribution.
- Trails the frontier on measured chat quality.
How good is it?
An open text model for chat and everyday questions, though it trails most models at multi-step tasks and changing course on request.
- multi-step work it carries out for youArena Agent · 43rd of 55
- changing course when you give new instructionsArena Agent · Steerability · 47th of 55
EverydayGeneral questions and everyday reasoning
Arena Text (overall)60th of 168 · 1440
CodingWriting and fixing code on its own
Arena Coding56th of 168 · 1495
AgenticPlanning, calling tools, staying on task
Arena Agent43rd of 55 · −0.066
WritingDrafting and rewriting prose
Arena Creative Writing60th of 168 · 1406
Placings on Arena's agent boards, from live sessions people ran themselves. A model can lead on one of these and sit mid-field on the others.
Boards this model appears on that none of the ratings above are built on.
Every published score for this model20 scoresEvery figure we hold, from 20 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Apple M1 Pro (16-core GPU) · 32 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Comfortable fit
Apple M3 Ultra (80-core GPU) · 512 GB
Room to spare. 105.3 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked between 4 hours and 28 hours ago — each listing carries its own date.
Cheapest of the 3 listings we can compare like for like — at 1M of context, out of 14 in the table below. 3 cheaper rows there are outside that comparison: a different context length or a different quantisation.
- per 1M tokens
- $0.30 in / $1.20 out
- Context served
- 1M
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| CoreWeavefp4Through OpenRouter | $0.23 / $0.96checked 4 hours ago | 262K236K max reply | 17 tok/s | No | No | Confirmed |
| GMICloudfp8Through OpenRouter | $0.24 / $0.96checked 4 hours ago | 1M524K max reply | 35 tok/s | No | Yesunknown period | Unknown |
| DeepInfrafp8Direct and through OpenRouter | $0.28 / $1.10checked 4 hours ago | 524K131K max reply through OpenRouter | 20 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| OpenRouterOpenRouter's own listing | $0.30 / $1.20checked 4 hours ago | 1M | not measured | Unknown | Unknown | Unknown |
| Novita AIfp8Direct and through OpenRouter | $0.30 / $1.20checked 4 hours ago | 1M131K max reply through OpenRouter | 53 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| AtlasCloudfp8Through OpenRouter | $0.30 / $1.20checked 4 hours ago | 524K524K max reply | 68 tok/s | No | Yesunknown period | Unknown |
| Parasailfp8Through OpenRouter | $0.30 / $1.20checked 4 hours ago | 1M524K max reply | 86 tok/s | No | No | Confirmed |
| StreamLakefp8Through OpenRouter | $0.30 / $1.20checked 4 hours ago | 1M512K max reply | 42 tok/s | No | Yesunknown period | Unknown |
| Minimaxfp8Through OpenRouter | $0.30 / $1.20checked 4 hours ago | 524K512K max reply | 53 tok/s | No | Yesunknown period | Unknown |
| Together AIThrough OpenRouter | $0.30 / $1.20checked 4 hours ago | 524K472K max reply | 44 tok/s | No | No | Confirmed |
| Venice AIfp8Through OpenRouter | $0.30 / $1.20checked 4 hours ago | 524K66K max reply | 20 tok/s | No | No | Confirmed |
| SambaNovaDirect and through OpenRouter | $0.60 / $2.40checked 4 hours ago | 1M944K max reply through OpenRouter | 69 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| MaraThrough OpenRouter | $0.60 / $2.40checked 28 hours ago | 1M944K max reply | 120 tok/s | No | No | Confirmed |
| ModelRunfp4Through OpenRouter | $0.75 / $3.00checked 10 hours ago | 1M944K max reply | 88 tok/s | No | No | Confirmed |
DeepInfra
Novita AI
Together AI
SambaNovaAcross the 14 listings we hold: 13 say they do not train on prompts (3 of them only through OpenRouter), 0 say they do and 1 does not say. 9 appear in the zero-retention registry we check (3 of them only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| CoreWeavefp4Through OpenRouter | ✓ | ✓ | ✓ |
| GMICloudfp8Through OpenRouter | ✓ | ✓ | ✗ |
| DeepInfrafp8Direct and through OpenRouter | ✓ | ✓ | ✗ |
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| Novita AIfp8Direct and through OpenRouter | ✓ | ✗ | ✗ |
| AtlasCloudfp8Through OpenRouter | ✓ | ✗ | ✗ |
| Parasailfp8Through OpenRouter | ✓ | ✗ | ✗ |
| StreamLakefp8Through OpenRouter | ✗ | ✓ | ✗ |
| Minimaxfp8Through OpenRouter | ✓ | ✓ | ✗ |
| Together AIThrough OpenRouter | ✓ | ✓ | ✓ |
| Venice AIfp8Through OpenRouter | ✓ | ✗ | ✗ |
| SambaNovaDirect and through OpenRouter | ✓ | ✓ | ✗ |
| MaraThrough OpenRouter | ✓ | ✓ | ✗ |
| ModelRunfp4Through OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 13 of 14 listings say yes, 1 says no. JSON output: 10 of 14 listings say yes, 4 say no. Strict schema: 4 of 14 listings say yes, 10 say no.
Models people weigh against MiniMax M3
When we formed this view
Recent changes
Each date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- We do not hold the active parameter count for it, so how much of it runs on any one token is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 14 listings does not say whether it trains on prompts, and 3 answer only through OpenRouter, not for their own listing.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsCustom licence, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Custom licence
This model ships custom license terms that don't map to a known template. We haven't parsed them, so commercial use, redistribution and derivatives are unverified — review the original terms before shipping.
Identifiers
- Hugging Face
- MiniMaxAI/Minimax-M3
- Architecture
- Mixture of experts
- Takes in, gives back
- Text, images and video in, text out
- Catalogue slug
- minimax-minimax-m3