What is MiniMax M3?
MiniMax M3 is MiniMax’s open weights model for open-weight multimodal agents, long-context coding, browsing and private deployment experiments. It was released June 1, 2026. This guide separates documented specifications from vendor benchmark claims and gives teams a practical way to decide whether the model belongs in a real evaluation.
QUICK VERDICT: MiniMax M3 is compelling for teams seeking open weights, native multimodality and long context in one model. The practical question is whether its harness and serving stack match the provider demonstrations.
Field | Verified value |
|---|---|
Provider | MiniMax |
Release date | June 1, 2026 |
Availability | Available; open-weight release announced |
License | Open weights |
Context window | Up to 1 million tokens; minimum 512K guaranteed by API documentation |
Maximum output | Shared within the total context limit; check the API model table for current request caps |
Modalities | Native text and visual understanding; text output through the language-model API |
API pricing | Standard API up to 512K: $0.30 input, $0.06 cache read and $1.20 output per million tokens; rates double above 512K |
Access | MiniMax API, MiniMax Code, Token Plan, Hugging Face and planned local deployment resources |
Best fit | open-weight multimodal agents, long-context coding, browsing and private deployment experiments |
Last verified | September 2, 2026 |
What changed with MiniMax M3
MiniMax M3 should not be evaluated as a name change. The important differences are the model’s reasoning controls, tool behavior, context policy, access route and economic profile. Those details determine whether a benchmark result transfers to production.
Area | What changed or matters |
|---|---|
Attention | MiniMax Sparse Attention supports long context |
Multimodality | Trained with multimodal data from the start |
Agents | Designed for autonomous decomposition, tool use and multi-step work |
Open deployment | MiniMax announced Hugging Face, GitHub and private-cluster support |
The correct comparison baseline is MiniMax M2.7. Teams planning a new deployment should also include MiniMax M3 where access allows. Testing only the newest model or only the incumbent hides the migration cost, output-style changes and tool-use regressions that often matter more than a small benchmark gap.
MiniMax M3 capabilities
The clearest fit is open-weight multimodal agents, long-context coding, browsing and private deployment experiments. That does not mean every task in those categories should use the model. A production system combines the base model with prompts, retrieval, tools, permissions, memory, retry logic and human review. The model card describes only one layer of that system.
Capability | Practical implication |
|---|---|
Long context | Up to 1 million tokens; minimum 512K guaranteed by API documentation. Test retrieval accuracy at realistic lengths rather than assuming every token receives equal attention. |
Output capacity | Shared within the total context limit; check the API model table for current request caps. Long output is useful only when the verification process can keep up. |
Modalities | Native text and visual understanding; text output through the language-model API. Confirm format-specific accuracy with your own files. |
Agent use | Use explicit tool schemas, narrow permissions, approval gates and recoverable operations. |
Reasoning | Record the exact effort level because cost, latency and quality can change materially. |
For coding, evaluate repository navigation, test creation, regression rate, review burden and recovery after a failed tool call. For research, measure citation correctness, source coverage and whether the model distinguishes evidence from inference. For computer or browser use, record every unsafe click, wrong field, lost state and unapproved action.
Benchmarks and evidence
The evidence below comes from MiniMax or named launch partners. It is useful for identifying intended strengths, but it is not equivalent to an independent head-to-head test. Prompts, tools, reasoning budgets, sampling, infrastructure and scoring rules can differ.
Evidence | Reported result | How to interpret it |
|---|---|---|
BrowseComp | 83.5 reported by MiniMax | Vendor-reported benchmark |
Paper reproduction | Nearly 12-hour run with 18 commits and 23 figures | Provider demonstration |
Kernel optimization | 147 iterations and 9.4 times speedup on a Hopper GPU task | Provider demonstration with a specific environment |
Treat benchmark results as a shortlist signal. Before purchasing or migrating, reproduce representative work under one frozen protocol. Keep model snapshots, reasoning effort, tool access and token budgets constant. Report repeated-run pass rates and total cost per accepted result, not a single best attempt.
Pricing, access and deployment
MiniMax API, MiniMax Code, Token Plan, Hugging Face and planned local deployment resources. The current pricing reference is Standard API up to 512K: $0.30 input, $0.06 cache read and $1.20 output per million tokens; rates double above 512K. Provider pricing changes frequently and can include cache rates, batch discounts, regional premiums, long-context multipliers, priority processing and tool-call fees. Recheck the official pricing page before budgeting.
Cost driver | What to measure |
|---|---|
Input tokens | Prompt, retrieved context, tool results and repeated history |
Output tokens | Visible answer plus any billable reasoning or generated artifacts |
Caching | Eligible repeated prefixes, cache-read price and expiration policy |
Tools | Search, computer use, code execution and third-party API fees |
Retries | Failed runs, verifier loops and human rework |
Success-adjusted cost | Total spend divided by deliverables that pass review |
A cheaper token price can lose to a more expensive model if it takes more steps, retries more often or produces work that needs heavy correction. Conversely, a frontier model can be wasteful when a smaller model already passes the task rubric. Route by measured task difficulty rather than brand prestige.
MiniMax M3 limitations
- Rates double above 512K input tokens.
- Provider demonstrations depend on tools, environments and long budgets.
- Open-weight serving at full context needs substantial infrastructure.
- Native multimodality still needs format-specific testing.
- The release status of every weight and fine-tuning artifact should be checked before procurement.
High-stakes medical, legal, financial, security and scientific work requires qualified review. Store the exact model identifier, prompt version, tools, source documents and approvals for each consequential run. Build a rollback path before granting write access to repositories, browsers, databases or cloud infrastructure.
Risk | Minimum control |
|---|---|
Hallucination | Require source checks or executable tests |
Prompt injection | Separate untrusted content from instructions and restrict tools |
Over-permission | Use least privilege and approval gates |
Silent model change | Pin snapshots where possible and run regression tests |
Data exposure | Review retention, regional processing and provider terms |
Runaway cost | Set token, time, tool-call and retry budgets |
How to evaluate MiniMax M3
Create 20 to 50 tasks from real work. Freeze the tasks and rubric before testing. Include easy tasks, normal tasks, edge cases and adversarial inputs. Give each model equivalent tools and enough budget to finish, but cap time and retries. Repeat non-deterministic runs so one lucky result does not decide the winner.
Test area | Record |
|---|---|
Task completion | Pass or fail plus rubric score |
Reliability | Repeated-run success and variance |
Quality | Factuality, instruction adherence and usefulness |
Tool use | Wrong calls, retries, recovery and permission errors |
Efficiency | Wall time, tokens, cache use, tool fees and human review |
Safety | Unsafe actions, injection response and sensitive-data handling |
Migration | Prompt changes, integration work and output-style regressions |
Compare MiniMax M3 with MiniMax M2.7 and MiniMax M3. Choose the least expensive configuration that meets the acceptance threshold with an adequate safety margin. Re-run the suite after any model snapshot, system prompt, retrieval or tool change.
Frequently asked questions
Is MiniMax M3 available now?
Available; open-weight release announced. The documented access routes are MiniMax API, MiniMax Code, Token Plan, Hugging Face and planned local deployment resources. Availability can differ by region, plan and partner platform.
Is MiniMax M3 open source?
MiniMax M3 has an open-weight route. Open weights do not necessarily mean the training data, training code and full software stack are open source. Review the exact license before commercial deployment.
How much does MiniMax M3 cost?
Standard API up to 512K: $0.30 input, $0.06 cache read and $1.20 output per million tokens; rates double above 512K. Budget with measured end-to-end workloads because token rates alone omit retries, tools, caching, long-context multipliers and review time.
What is MiniMax M3 best used for?
Its strongest documented fit is open-weight multimodal agents, long-context coding, browsing and private deployment experiments. Start with a supervised pilot and keep human sign-off for consequential work.
Should I migrate from MiniMax M2.7?
Only after a side-by-side evaluation. Measure pass rate, total cost, latency, output style, tool reliability and the engineering work required to migrate. A newer model is not automatically the better operational choice.
Related model guides
Start with the broader Best AI Models directory.
Compare with the Kimi K3 guide.
Compare with the Qwen 3.8 guide.
Official sources and update policy
MiniMax pay-as-you-go pricing.
Checked September 2, 2026. We update this guide when the provider changes the model specification, pricing, access, licensing or safety documentation. Vendor benchmarks are attributed and are not presented as independent testing. No vendor payment or affiliate relationship determined inclusion.