MiniMax M1 80k
MiniMax M1 80k is a chat model from MiniMax, released as part of the M1 series where the numeric suffix distinguishes the extended-generation variant from its shorter sibling.
API Pricing
| Provider | Input / 1M | Output / 1M |
|---|---|---|
| $0.550 | $2.20 |
Prices updated daily. Last check: Oct 18, 2026
Compare API pricing for every MiniMax model →MiniMax M1 80k pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- MMLU-Pro81.6%
- GPQA Diamond69.7%
- Humanity's Last Exam8.9%
Coding
- LiveCodeBench71.1%
Math
- AIME 202561.0%
- AIME84.7%
- MATH-50098.0%
Agentic & Tool Use
- Terminal-Bench Hard3.0%
- τ²-bench34.2%
Instruction & Long Context
- IFBench41.8%
- Long-Context Reasoning57.7%
Benchmarks measured Oct 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- MiniMax
- Modalities
- Text
Capabilities
- Open Source
- Yes
Strengths & Limitations
Strengths
- Configured for the longer of the two M1 generation budgets, giving more room for extended intermediate reasoning than the 40k sibling
- Explicit variant naming makes it easy to route traffic between the 40k and 80k configurations based on task difficulty
- Reasoning-style generation suits multi-step math, constraint-heavy coding, and analytical tasks over direct-answer models
- Comes from MiniMax, giving buyers a non-US-origin option when comparing reasoning model providers
- Available through multiple inference providers, so the pricing table on this page can be used to compare cost across hosts
- Sits in a family with a documented cheaper sibling, which simplifies A/B testing of quality versus spend
Limitations
- Reasoning-style generation produces high output token counts, which raises both latency and per-request cost relative to direct-answer models
- Our database does not carry verified context window, modality, or tool-calling data for this listing
- No populated throughput or time-to-first-token benchmarks are available in our data, so speed cannot be compared numerically here
- The 80k configuration is the more expensive of the two M1 variants to run for a given prompt
- Serving configurations differ between providers, so behavior and effective limits may not be identical across hosts
Key Features
About MiniMax M1 80k
Common Use Cases
MiniMax M1 80k fits workloads where the model needs to work through a problem rather than answer immediately: competition-style math, algorithmic coding problems that must satisfy multiple constraints, technical troubleshooting, and analytical writing where the intermediate steps matter to the final quality. The 80k configuration is the one to reach for on the hardest subset of those prompts — the 40k variant is the more economical default for routine questions. It is a poor fit for high-volume, latency-sensitive tasks such as classification, autocomplete, short-form chat, or anything where you are paying per output token at scale, since reasoning-style generation inflates output length substantially. Teams evaluating it typically benchmark it head-to-head against other reasoning models on their own task set and compare cost per completed answer, not cost per token, using the pricing table on this page.
Frequently Asked Questions
What does the "80k" in MiniMax M1 80k refer to?
It denotes the larger generation budget configuration of the MiniMax M1 series. MiniMax released the family with two settings, and the 80k variant allows a longer stretch of intermediate reasoning output before the final answer than the 40k variant does.
How much does MiniMax M1 80k cost to run?
Pricing varies by provider and by pricing type — input tokens, output tokens, and any caching or batch discounts are billed separately, and different hosts quote different rates for the same weights. See the pricing table on this page for current figures. Note that reasoning-style models emit far more output tokens than direct-answer models, so output token pricing tends to dominate the real bill.
What is MiniMax M1 80k best used for?
Tasks that benefit from extended step-by-step work: multi-step mathematics, constraint-heavy code generation, technical analysis, and problem solving where intermediate reasoning improves the final answer. It is less suited to high-throughput classification, short chat, or latency-critical interactive features.
Should I use the 80k or the 40k variant?
Use the 40k configuration as the default for routine prompts and reserve the 80k configuration for the harder subset where the shorter budget produces incomplete or incorrect reasoning. Because the two are otherwise the same family, routing between them by task difficulty is a straightforward way to control spend.
How fast is MiniMax M1 80k?
We do not have populated output-tokens-per-second or time-to-first-token measurements for this listing, so we cannot give a numeric comparison. As a general point, reasoning models feel slower end-to-end than their raw token rate suggests because a large share of generated tokens are intermediate reasoning rather than the visible answer.