MiniMax
SG
Latency
-
Throughput
41tps
Uptime
Total Context
128K
Max Output
-
Status
Standby
Pricing
Input Price
Provider priced
/M tokens
Output Price
Provider priced
/M tokens
Cache Creation (5min)
-
/M tokens
Cache Read (5min)
-
/M tokens
M
MiniMax: MiniMax-M2.5
minimax/minimax-m2.5
Current version2.5Modestable
响应轻快，适合日常对话、快速整理与轻量写作
Modalities
Input
Provider priced per 1M
Output
Provider priced per 1M
context
128K
Weekly Tokens
Live catalog
Released
Live catalog
Overview for MiniMax-M2.5
A compact summary of this model's best-fit workloads, routing status and production integration shape.
Architecture
MiniMax OpenAI-compatible route
Model design signal
Traffic signal
Live catalog
Recent public usage volume
Benchmark score
-
Composite coding and reasoning signal
Best for
tool usetext generation
Model tags
chatstreamingtools
Routing snapshot
Primary provider
MiniMax
Provider region
Provider policy
Current version
2.5
Routing health
Missing key
Missing MINIMAX provider key
MiniMax
Fallback route
Context and limits
Total Context
128K
Max prompt
128K
Max Output
-
Pricing tier
x0.5 gateway rate
Pricing snapshot
Input PriceProvider priced
Output PriceProvider priced
Cache creation-
Cache read-
Integration checklist
✓OpenAI-compatible chat completions
✓Call with model ID minimax/minimax-m2.5
✓Use base URL https://ethergate.cn/v1
✓Create a workspace-scoped API key
Capability profile
chatStreamingTool calling
MiniMax policies apply for retention, training and regional handling.
Parameter support
max_tokenstemperaturetop_ptoolstool_choiceCheck long-context pricing once prompts exceed 128K.
Very large prompts can add latency; test production payloads before launch.
Providers for MiniMax-M2.5
EtherGate routes requests to configured providers that can handle your prompt size and parameters, with fallbacks to maximize uptime. uptime.
Performance for MiniMax-M2.5
Compare different providers across EtherGate
Throughput
30dMiniMaxAverage 40 tok/s
Latency
30dMiniMaxAverage 1.84 s
E2E Latency
30dMiniMaxAverage 12.54 s
Tool Call Error Rate
30dMiniMaxAverage 4.83 %
Structured Output Error Rate
30dMiniMaxAverage 1.38 %
Effective Pricing for MiniMax-M2.5
Actual cost per million tokens across providers over the past hour
Weighted Average
Weighted Avg Input Price
Provider priced
Weighted Avg Output Price
Provider priced
| Provider | Input Price | Output Price | Cache Hit Rate |
|---|---|---|---|
| MiniMax | Provider priced | Provider priced | 42% |
Input Price / 1M tokens
7dMay 12May 14May 16May 18
Output Price / 1M tokens
7dMay 12May 14May 16May 18
Benchmarks for MiniMax-M2.5
Performance metrics and benchmarks
Ranking Distribution
1617
#1
1653
#2
1322
#3
1368
#4
Category Performance
Composite coding and reasoning signal
Score
-
Models Arena
3D
1297
53.0% Win · 170.0s Avg
ASCII
1196
50.0% Win · 152.0s Avg
Code Categories
1288
53.4% Win · 254.7s Avg
Data Visualization
1283
53.3% Win · 236.9s Avg
Game Development
1292
54.8% Win · 270.4s Avg
Image to HTML
1205
46.7% Win · 116.4s Avg
SVG
1239
55.2% Win · 133.4s Avg
UI Component
1303
54.5% Win · 252.7s Avg
Website
1266
51.8% Win · 306.2s Avg
Apps using MiniMax-M2.5
Top public apps this month
Missing key: Missing MINIMAX provider key
1
Hermes Agent
Open-source agent workspace for model routing
353B
Tokens
2
OpenClaw
AI coding agent connected to EtherGate
116B
Tokens
3
AA-LCR Benchmark
Applied research lab for frontier models
35.9B
Tokens
4
Claude Code Bridge
Developer tool using OpenAI-compatible APIs
26.8B
Tokens
5
Roo Code
VS Code extension for model-assisted coding
17.3B
Tokens
Monthly token usage
Apr 20May 1May 10May 18
Recent activity on MiniMax-M2.5
Total usage per day on EtherGate
Prompt tokensReasoning tokensCompletion tokens
Uptime stats for MiniMax-M2.5
Uptime stats for MiniMax-M2.5 across configured providers
Average uptime
Not configured
Sample code and API for MiniMax-M2.5
EtherGate normalizes requests and responses across providers for you.
EtherGate supports reasoning-enabled models that can return step-by-step reasoning metadata. Use the reasoning parameter when a provider supports it, and preserve reasoning_details when continuing a conversation. Learn more about reasoning tokens.
Request
Use the OpenAI-compatible Chat Completions endpoint. Keep your SDK, change the base URL, API key and model ID.
OpenAI SDK
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.ETHERGATE_API_KEY,
baseURL: "https://ethergate.cn/v1",
});
const completion = await client.chat.completions.create({
model: "minimax/minimax-m2.5",
messages: [{ role: "user", content: "Explain your best use case." }],
});Tool configuration examples
Use this model ID directly in agent tools that accept OpenAI-compatible endpoints.
QClaw macOS/Linux
curl -fsSL 'https://ethergate.cn/console/skills/ethergate-provider/install.sh' | bash -s -- --target qclaw --api-key '<your-ethergate-key>' --locale 'en'Required headers
Authorization
Bearer $ETHERGATE_API_KEYHeader
Content-Type: application/jsonBase URL
https://ethergate.cn/v1Optional headers
Optional attribution headers help you identify traffic by app or website when debugging usage.
HTTP-Refererhttps://your-app.example
X-TitleYour app name
Using third-party SDKs
openai-js
openai-python
Vercel AI SDK
LangChain