Requesty
DeepInfra Inc. logo
DeepInfra Inc.

MiniMaxAI/MiniMax-M2.5

MiniMax M2.5 is SOTA in coding, agentic tool use and search, office work, and a range of other economically valuable tasks, boasting scores of 80.2% in SWE-Bench Verified, 51.3% in Multi-SWE-Bench, and 76.3% in BrowseComp (with context management).

ReasoningTool callingCachingJSON schema
Try MiniMaxAI/MiniMax-M2.5CompareDocs

Specifications

Context window197K tokens
Max outputβ€”
API typechat
AddedMay 27, 2026
Model IDdeepinfra/MiniMaxAI/MiniMax-M2.5
Data retentionNo
Used for trainingNo
Provider locationπŸ‡ΊπŸ‡Έ US
Privacy policyDeepInfra Privacy Policy→

Benchmarks

Released 2026-02-12
Coding Indexcoding
37.4%

Artificial Analysis Coding Index β€” a composite of coding evaluations including LiveCodeBench, SciCode and Terminal-Bench.

GPQA Diamondreasoning
84.8%

Graduate-level physics, chemistry & biology questions designed to resist Googling.

Intelligence Indexreasoning
41.9%

Artificial Analysis Intelligence Index β€” a composite of multiple evaluations measuring overall model capability.

Scores are sourced from official model cards, Artificial Analysis, and public leaderboards. Benchmarks measure specific skills and do not capture every aspect of model quality β€” always test on your own workload.

Pricing

Input / 1M
$0.15
Output / 1M
$1.15
Cache write
β€”
Cache read / 1M
$0.03
Estimated cost
100K input + 10K output$0.0265
1M input + 100K output$0.27
10M input + 1M output$2.65

Requesty charges exactly what the upstream provider charges β€” no markup, no per-request fees. Prompt caching and smart routing can reduce effective cost by 30-80%.

Quickstart

Drop-in compatible with the OpenAI SDK. Change the base URL, swap in your Requesty API key, and set the model to deepinfra/MiniMaxAI/MiniMax-M2.5.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_REQUESTY_API_KEY",
    base_url="https://router.requesty.ai/v1",
)

response = client.chat.completions.create(
    model="deepinfra/MiniMaxAI/MiniMax-M2.5",
    messages=[
        {"role": "user", "content": "Explain quantum computing in one paragraph."},
    ],
)

print(response.choices[0].message.content)

Other DeepInfra Inc. models

zai-org/GLM-5.1
203K ctx$1.05/1M
Coding 43
nvidia/Nemotron-3-Nano-30B-A3B
262K ctx$0.05/1M
nvidia/NVIDIA-Nemotron-3-Super-120B-A12B
262K ctx$0.10/1M
Coding 31
moonshotai/Kimi-K2.6
262K ctx$0.75/1M
Coding 47
gemma-4-31B-it
262K ctx$0.13/1M
deepseek-ai/DeepSeek-V4-Pro
1.0M ctx$1.30/1M
Coding 48

Frequently asked questions

How much does MiniMaxAI/MiniMax-M2.5 cost?
MiniMaxAI/MiniMax-M2.5 is priced at $0.15 per million input tokens and $1.15 per million output tokens when accessed via Requesty. Prompt caching is supported, which can cut effective input cost by up to 90% on repeated context. Requesty charges exactly what the upstream provider charges β€” we don't add markup.
What is the context window of MiniMaxAI/MiniMax-M2.5?
MiniMaxAI/MiniMax-M2.5 has a context window of 197K tokens. That's roughly 262 words of input you can fit in a single prompt.
How does MiniMaxAI/MiniMax-M2.5 perform on benchmarks?
MiniMaxAI/MiniMax-M2.5 scores 95.3% on τ²-Bench, 84.8% on GPQA Diamond, 42.6% on SciCode. See the full benchmark chart above for results across MMLU Pro, GPQA Diamond, SWE-Bench Verified, HumanEval, MATH, AIME, MMMU, and LiveBench.
What can MiniMaxAI/MiniMax-M2.5 do?
MiniMaxAI/MiniMax-M2.5 supports tool calling, extended reasoning, prompt caching, structured outputs (JSON schema). You can call it through any OpenAI-compatible client by pointing base_url to Requesty.
How do I use MiniMaxAI/MiniMax-M2.5 with the OpenAI SDK?
Install the OpenAI SDK, set base_url to "https://router.requesty.ai/v1", set your API key to your Requesty key, and set the model to "deepinfra/MiniMaxAI/MiniMax-M2.5". The Quickstart above shows Python, JavaScript and cURL snippets.

Access MiniMaxAI/MiniMax-M2.5 through Requesty

One API key, 400+ models, OpenAI-compatible. No markup on provider prices, automatic failover, and smart caching built-in.

Get started freeAll DeepInfra Inc. models