How to Run MiniMax M3.1 Flash on Ollama: Cloud Setup Guide (2026)
MiniMax M3.1 Flash is a 428B-parameter cloud coding model on Ollama with a 976K context window. Run minimax-m3.1-flash:cloud and set up API access.
MiniMax M3.1 Flash is MiniMax's coding and agentic model, publicly launched October 19, 2026 after a preview that began September 27 under the internal codename "Space Bunny," which topped several leaderboards before MiniMax confirmed what it actually was. It's a mixture-of-experts model with 428 billion total parameters and roughly 23 billion active per token, built on MiniMax Sparse Attention (MSA), a sparse-attention design that makes its 976,000 token context window practical instead of prohibitively slow. The previous M2 generation topped out at 204,800 tokens, so this is close to a five-times jump in one release. On Ollama, the only listed tag is `minimax-m3.1-flash:cloud`, which sends your prompt to MiniMax's hosted infrastructure through Ollama's servers.
MiniMax has not released open-source weights for M3.1 Flash, and has stated weights are planned for October 23, 2026, one day after this guide's publish date. Until that happens, and until Ollama adds a corresponding local tag if it does, `:cloud` is the only way to run this model through Ollama. Check the model's library page on ollama.com for an updated tag list if you're reading this after that date.
This guide covers the cloud setup: installing Ollama, signing in, running `minimax-m3.1-flash:cloud` from the terminal, and generating an API key for scripts and agents. The alternatives section compares it to the base MiniMax M3 model, GLM 5.2, and DeepSeek R1 for other cloud and local options.
Prerequisites
- Ollama 0.12 or later, installed on Linux, macOS, or Windows (no GPU or high-RAM machine required for the cloud route)
- A free account at ollama.com for the `ollama signin` step
- A stable internet connection, since inference runs on MiniMax's and Ollama's servers, not your hardware
- Basic terminal familiarity for running `ollama run` and `curl` commands
- (Optional) An API key from ollama.com/settings/keys if you plan to call MiniMax M3.1 Flash from your own scripts or agents
Need more GPU power?
Rent a 4x A100 80GB on Vast.ai from $6.50/hr. On-demand GPU rentals by the hour, useful for running larger models without buying hardware.
In This Guide
What MiniMax M3.1 Flash Is and Why Ollama Runs It in the Cloud
MiniMax M3.1 Flash, internally model ID MiniMax-M3.1-Flash-Preview, is a mixture-of-experts coding and agentic model from MiniMax, a Chinese AI lab. It carries 428 billion total parameters with approximately 23 billion active per token, a native visual encoder for image and video input, and MiniMax Sparse Attention, which is what makes a 976,000 token context window usable at roughly 150 output tokens per second rather than collapsing under the cost of full dense attention at that length. Reasoning is always on, with depth set through an effort parameter: low, medium, high, xhigh, or max, defaulting to max.
| Spec | Value |
|---|---|
| Total parameters | 428 billion (MoE) |
| Active parameters | ~23 billion per token |
| Attention mechanism | MiniMax Sparse Attention (MSA) |
| Context window | 976,000 tokens |
| Input modalities | Text, image, video |
| Output modality | Text only |
| Reasoning effort levels | low, medium, high, xhigh, max (default: max) |
| Ollama tag | `minimax-m3.1-flash:cloud` (only option) |
M3.1 Flash is the successor to the M2 family, which topped out at a 204,800 token context window, making this release close to a five-times increase. MiniMax markets it as a generational leap in agentic tasks, coding, Computer Use, and cybersecurity work, and says online inference runs close to three times faster than its predecessor at roughly 70% lower usage cost. Treat those specific multipliers as MiniMax's own vendor claims rather than independently verified benchmark results, since no third-party evaluation numbers were available at the time of this guide's publication.
Through Ollama Cloud, pricing runs $0.15 per million input tokens, $0.015 per million cached input tokens, and $0.60 per million output tokens as of this guide's publish date. That pricing applies within Ollama's own usage limits, separate from billing through MiniMax's direct API, M Plan subscription, or MiniMax Code.
Set Up Ollama Cloud and Run MiniMax M3.1 Flash
Running MiniMax M3.1 Flash through Ollama takes three steps: install Ollama, sign in, and run the model. Nothing downloads a 428B-parameter file to your disk, since inference happens on MiniMax's infrastructure rather than locally.
Step 1: Install Ollama
# Linux and macOS, one-command installer
curl -fsSL https://ollama.com/install.sh | shOn Windows, download the installer from ollama.com/download, or use winget:
winget install Ollama.OllamaVerify the installation:
ollama --versionollama version is 0.12.3Step 2: Sign In to Ollama Cloud
ollama signinThis prints a sign-in URL and opens your browser. Create a free account at ollama.com, or log in if you already have one, then approve the device.
Signing in to ollama.com...
Signed in as your-usernameStep 3: Run MiniMax M3.1 Flash from the Terminal
ollama run minimax-m3.1-flash:cloudOllama fetches a small manifest rather than model weights, then drops you into a prompt:
pulling manifest
pulling 7d3ae912bc84... 100% ââââââââââââââââââ 4.2 KB
success
>>> Send a message (/? for help)Test it with an agentic-style prompt, since that's the model's core design target:
>>> Walk through debugging a Python script that works locally but fails in CI with a ModuleNotFoundError, and list the three most likely causes in order.The first response after signing in can take 10-20 seconds while Ollama establishes the cloud session. Responses after that stream back at normal speed.
Step 4: Verify the Model
ollama list`minimax-m3.1-flash:cloud` shows up at a few KB, not hundreds of gigabytes. That confirms it's the cloud passthrough entry, not a local download.
More how-to guides like this
Step-by-step AI guides and tool walkthroughs, straight to your inbox, twice a week.
Use MiniMax M3.1 Flash in Your Own Scripts and Agents
Beyond the interactive `ollama run` session, MiniMax M3.1 Flash's cloud tag works through Ollama's REST API and its built-in agent launch shortcuts, so existing tools that already talk to Ollama can switch to it with a one-line model name change.
Generate an API Key
Visit ollama.com/settings/keys while signed in, click "Create API key," and copy the value.
export OLLAMA_API_KEY=your_api_key_hereCall MiniMax M3.1 Flash from the Local Endpoint
curl http://localhost:11434/api/chat -d '{
"model": "minimax-m3.1-flash:cloud",
"messages": [
{ "role": "user", "content": "Review this pull request diff for race conditions in the retry logic." }
],
"stream": false
}'Python Example
from ollama import Client
client = Client(host="https://ollama.com", headers={"Authorization": "Bearer " + api_key})
response = client.chat(
model="minimax-m3.1-flash:cloud",
messages=[{"role": "user", "content": "Summarize this 400-page compliance document into a 5-point checklist."}],
)
print(response["message"]["content"])Launch Directly Into an Agent Scaffold
Ollama ships launch shortcuts that point an agent framework at a given model in one command:
ollama launch claude --model minimax-m3.1-flash:cloud
ollama launch opencode --model minimax-m3.1-flash:cloud
ollama launch hermes --model minimax-m3.1-flash:cloud
ollama launch openclaw --model minimax-m3.1-flash:cloudEach of these drops you straight into that agent's interface running on MiniMax M3.1 Flash. Reasoning effort control (low through max) is exposed on MiniMax's own API and M Plan; through Ollama's cloud passthrough, the model runs at its default max effort unless Ollama adds explicit support for that parameter on this tag.
Troubleshooting
`ollama pull minimax-m3.1-flash` fails with "pull model manifest: file does not exist"
Cause: Ollama's library only lists `minimax-m3.1-flash:cloud`, with no local or quantized tag
Fix: Use `ollama run minimax-m3.1-flash:cloud` with the `:cloud` suffix included. MiniMax has said open weights are planned for October 23, 2026, so check the library page again if that date has passed.
`ollama run minimax-m3.1-flash:cloud` returns "model not found"
Cause: The installed Ollama version predates cloud model support
Fix: Update Ollama by re-running the install command, then retry. Cloud models require Ollama 0.6.x or later.
"unauthorized" error or repeated sign-in prompts
Cause: The machine is not signed in, or the session expired
Fix: Run `ollama signin` again and complete the browser approval. Check ollama.com/settings/connections to confirm the device is listed as connected.
First response takes 10-20 seconds or longer
Cause: Cold start while Ollama establishes a session with MiniMax's cloud infrastructure
Fix: This is normal for the first request after signing in or after an idle period. Subsequent requests in the same session stream back at normal speed.
API requests to `https://ollama.com/api` return 401
Cause: Missing or invalid `OLLAMA_API_KEY`
Fix: Generate a new key at ollama.com/settings/keys and re-export the environment variable.
Video input in a prompt is ignored
Cause: The file path was not detected, often from a typo or an unsupported video format
Fix: Use an absolute path to a supported video file and confirm it exists before sending the prompt. Still images in JPEG or PNG are the more reliable fallback.
Alternatives to Consider
| Tool | Type | Price | Best For |
|---|---|---|---|
| MiniMax M3 | Cloud (Ollama) | Billed per token through Ollama Cloud | The base 428B-parameter M3 generation, also cloud-only through Ollama, for comparing against this Flash variant. |
| GLM 5.2 via Ollama Cloud | Cloud (Ollama) | Free within Ollama Cloud limits | A 744B cloud model with a 1M context window and effort-level control, also cloud-only through Ollama. |
| Qwen3-8 Flash Next | Local (Ollama) | Free, local install | A smaller, genuinely local flash-branded model if you want something you can actually pull and run today. |
| DeepSeek R1 | Local (Ollama) or VPS | Free | A true local reasoning model on hardware from 4GB (distilled) up to 64GB or more, with no cloud dependency. |
Frequently Asked Questions
Can I run MiniMax M3.1 Flash locally with Ollama?
Not yet. MiniMax has not released open weights for M3.1 Flash as of this guide's publish date, so there is no local tag to pull through Ollama or any other tool. MiniMax has stated weights are planned for October 23, 2026, one day after this guide was published.
Until weights ship, and until Ollama adds a corresponding local tag if it does, `minimax-m3.1-flash:cloud` through Ollama is the only way to use this model with Ollama's command syntax. Check ollama.com/library/minimax-m3.1-flash directly for the current tag list if you're reading this later.
How much does MiniMax M3.1 Flash cost through Ollama?
Through Ollama Cloud, usage is billed at $0.15 per million input tokens, $0.015 per million cached input tokens, and $0.60 per million output tokens as of this guide's publish date. `ollama signin` itself requires no payment information, but running prompts against `minimax-m3.1-flash:cloud` draws against Ollama's cloud usage billing.
MiniMax's own direct API, M Plan subscription, and MiniMax Code are separate billing paths from the Ollama Cloud route covered in this guide.
What was "Space Bunny" and how does it relate to MiniMax M3.1 Flash?
Space Bunny was the internal codename MiniMax used during M3.1 Flash's preview period, which began September 27, 2026. Under that name, the model topped several public leaderboards before MiniMax's October 19, 2026 announcement confirmed it was the preview build of M3.1 Flash. It's the same model, not a separate release.
How does MiniMax M3.1 Flash compare to the base MiniMax M3?
Attention and timing, mostly. Both carry 428 billion total parameters, but M3.1 Flash adds MiniMax Sparse Attention on top of a roughly 23 billion active-per-token profile, which is what makes its 976,000 token context window usable rather than painfully slow, and it's the newer release, publicly launched October 19, 2026.
Both are cloud-only through Ollama as of this guide's publish date, so the practical difference for an Ollama user is picking whichever tag reflects the newer architecture and longer context window, which is M3.1 Flash.
Does MiniMax M3.1 Flash support video input?
Yes. MiniMax M3.1 Flash accepts text, image, and video input, producing text-only output. Through Ollama, drop a file path into your prompt and Ollama attaches it automatically, the same mechanism used for image input with other multimodal models in Ollama's library.
What are the reasoning effort levels in MiniMax M3.1 Flash?
MiniMax exposes five effort levels on this model: low, medium, high, xhigh, and max, with max as the default. Higher effort levels trade response speed for deeper reasoning on complex tasks. This control is a feature of MiniMax's own API and M Plan; through Ollama's cloud passthrough, the model runs at its default max effort unless Ollama adds explicit support for that parameter on this tag.
How much faster is MiniMax M3.1 Flash than its predecessor?
MiniMax states that online inference runs close to three times faster than the prior generation, at roughly 70% lower usage cost, with output around 150 tokens per second depending on load. These are MiniMax's own figures. No independent third-party benchmark confirmation was available at the time this guide was published, so treat the specific multipliers as a vendor claim rather than a verified result.
Can I use MiniMax M3.1 Flash with an agent like Hermes Agent or OpenClaw?
Yes, and Ollama provides a shortcut for it. Running `ollama launch hermes --model minimax-m3.1-flash:cloud` or `ollama launch openclaw --model minimax-m3.1-flash:cloud` drops you directly into that agent's interface running on M3.1 Flash, without manually editing a config file first.
You can also configure Hermes Agent or OpenClaw manually by pointing their model configuration at Ollama's OpenAI-compatible endpoint, `http://localhost:11434/v1`, with the model name set to `minimax-m3.1-flash:cloud`.
Do I need a VPS or GPU to use MiniMax M3.1 Flash through Ollama?
No. Inference for the `:cloud` tag runs on MiniMax's infrastructure regardless of where you run `ollama`, so a laptop or desktop with no GPU is enough for the cloud route covered in this guide. There is currently no local install path for this model at all, since open weights have not been released.
Related Guides
How to Run MiniMax M3 on Ollama: Cloud Setup Guide (2026)
How to Run Muse Spark Open on Ollama: Cloud Setup Guide (2026)
How to Run GLM 5.2 on Ollama: Cloud Setup Guide (2026)
How to Run Kimi K2 on Ollama: Cloud Setup Guide (2026)
How to Run DeepSeek R1 Locally with Ollama (2026 Guide)
Best Local LLM Models to Run in 2026 (Benchmarks + Use Cases)
How to Run Qwen3.8-Flash-Next Locally with Ollama (2026 Guide)
How to Install Hermes Agent with Ollama Local Models (2026)