Skip to content
Updated minutes ago
MiniMax

MiniMax M2.7

MiniMax · moe · 456B parameters · 1,048,576 context

Quality
82.0

Parameters

456B

Context Window

1024K tokens

Architecture

MoE

Best GPU

B200 NVL (pair)

Cheapest API

$2.80/M

Quality Score

82/100

Intelligence Brief

MiniMax M2.7 is a 456B parameter Mixture-of-Experts (0 experts, N/A active) model from MiniMax, featuring Grouped Query Attention (GQA) with 80 layers and 6,144 hidden dimensions. With a 1,048,576 token context window, it supports tools, vision, structured output, code, math, multilingual, reasoning. On standardized benchmarks, it achieves MMLU 88, HumanEval 68, GSM8K 92. The most cost-effective API deployment is via minimax at $2.80/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $19929/month.

Architecture Details

TypeMOE
Total Parameters456B
Active Parameters45.9B
Layers80
Hidden Dimension6,144
Attention Heads48
KV Heads8
Head Dimension128
Vocab Size200,064

Memory Requirements

BF16 Weights

912.0 GB

FP8 Weights

456.0 GB

INT4 Weights

228.0 GB

KV-Cache per Token131072 bytes
Activation Estimate4.00 GB

Fits on (single GPU) — most practical first

GPU Compatibility Matrix

MiniMax M2.7 is compatible with 2% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
B300288GB
B100 SXM192GB
GB200 NVL72 (per GPU)192GB
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
H200 SXM141GB
H2096GB
GH20096GB
Ada Lovelace(11 GPUs)
L40S48GB
L4048GB
RTX 6000 Ada48GB
L2048GB
Ampere(16 GPUs)
A100 80GB SXM80GB
A100 80GB PCIe80GB
A1664GB
RTX A600048GB
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

B200 NVL (pair)optimal

FP8 · 2 GPUs · tensorrt-llm

100/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$19929

Cost/M Tokens

$27.08

Use this config →
B200 SXMoptimal

FP8 · 4 GPUs · tensorrt-llm

98/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$17044

Cost/M Tokens

$23.16

Use this config →
B100 SXMoptimal

FP8 · 4 GPUs · tensorrt-llm

98/100

score

Throughput

280.0 tok/s

Latency (ITL)

3.6ms

Est. TTFT

1ms

Cost/Month

$17082

Cost/M Tokens

$23.21

Use this config →

Deployment Options

API

API Deployment

minimax

$2.80/M

output tokens

Self-Hosted

Single GPU

Requires multi-GPU setup (456 GB VRAM needed)

Scale

Multi-GPU

B200 NVL (pair) x2

280.0 tok/s

TP· $19929/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
minimax$0.70$2.80
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
minimaxBest Value$0.70$2.80$18

Cost per 1,000 Requests

Short (500 tok)

$0.91

via minimax

Medium (2K tok)

$3.64

via minimax

Long (8K tok)

$11.20

via minimax

Performance Estimates

Throughput by GPU

B200 NVL (pair)
280.0 tok/s
B200 SXM
280.0 tok/s
B100 SXM
280.0 tok/s

VRAM Breakdown (B200 NVL (pair), FP8)

Weights
Weights 228.0 GBKV-Cache 2.7 GBActivations 32.0 GBOverhead 11.4 GB

Precision Impact

bf16

456.0 GB

weights/GPU

fp8

228.0 GB

weights/GPU

~280.0 tok/s

int4

114.0 GB

weights/GPU

Quality Benchmarks

Top 10%
91th percentile across all models
MMLU
88.0
Above Average (87th pctile)
HumanEval
68.0
Above Average (84th pctile)
GSM8K
92.0
Above Average (77th pctile)
MT-Bench
88.0
Bottom 25% (0th pctile)

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmsglang

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy MiniMax M2.7

Similar Models

MiniMax-Text-01

456B params · moe

Quality: 50

from $5.00/M

Lower quality, More expensiveCompare →
Smaller context, Lower quality, CheaperCompare →

Llama 3.1 405B

405B params · dense

Quality: 81

from $3.00/M

Smaller contextCompare →

Llama 4 Maverick

400B params · moe

Quality: 84

from $1.80/M

Smaller context, Lower quality, More expensiveCompare →

Frequently Asked Questions

How much VRAM does MiniMax M2.7 need for inference?

MiniMax M2.7 requires approximately 912.0 GB of VRAM at BF16 precision, 456.0 GB at FP8, or 228.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (131072 bytes per token) and activations (~4.00 GB).

What is the best GPU for MiniMax M2.7?

The top recommended GPU for MiniMax M2.7 is the B200 NVL (pair) (x2) using FP8 precision. It achieves approximately 280.0 tokens/sec at an estimated cost of $19929/month ($27.08/M tokens). Score: 100/100.

How much does MiniMax M2.7 inference cost?

MiniMax M2.7 API inference starts from $0.70/M input tokens and $2.80/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.