MiniMax M3
MiniMax · moe · 427B parameters · 1,048,576 context
Parameters
427B
Context Window
1024K tokens
Architecture
MoE
Best GPU
B200 NVL (pair)
Cheapest API
$1.20/M
Intelligence Brief
MiniMax M3 is a 427B parameter Mixture-of-Experts (128 experts, 4 active) model from MiniMax, featuring Grouped Query Attention (GQA) with 60 layers and 6,144 hidden dimensions. With a 1,048,576 token context window, it supports tools, vision, code. The most cost-effective API deployment is via openrouter at $1.20/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $19929/month.
Provider pricing
1 provider · canonical: openrouter| Provider | Input $/M | Output $/M ▲ | Notes |
|---|---|---|---|
| openroutercanonical | $0.300 | $1.20 | cheapest input · cheapest output |
Prices update via the nightly pricing cron + admin approvals at /admin/ingest-queue. The leaderboard's Input/Output cells show the canonical rate above; this table shows the full spread.
Recent changes
Loading…
Related models
5 suggestions- MiniMax-Text-01MiniMax · 45.9B$5.00/M out
- MiniMax M2.7MiniMax · 11B$1.20/M out
- MiniMax M2.7MiniMax M · 45.9Bfree/M out
Kimi K2.7 CodeKimi · 32.9B$3.50/M out
Llama 4 ScoutLlama 4 · 17B$0.300/M out
Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.
Architecture Details
Memory Requirements
BF16 Weights
795.4 GB
FP8 Weights
397.7 GB
INT4 Weights
198.9 GB
Fits on (single GPU) — most practical first
Fits on (multi-GPU with Tensor Parallelism)
Multi-GPU configurations use Tensor Parallelism (TP) to split model layers across GPUs. Requires NVLink or NVSwitch interconnect for optimal performance.
GPU Compatibility Matrix
MiniMax M3 is compatible with 2% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
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
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
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
Deployment Options
API Deployment
openrouter
$1.20/M
output tokens
Single GPU
Requires multi-GPU setup (398 GB VRAM needed)
Multi-GPU
B200 NVL (pair) x2
280.0 tok/s
TP· $19929/mo
API Pricing Comparison
| Provider | Input $/M | Output $/M | Badges |
|---|---|---|---|
| openrouter | $0.30 | $1.20 | Cheapest |
Cost Analysis
| Provider | Input $/M | Output $/M | ~Monthly Cost |
|---|---|---|---|
| openrouterBest Value | $0.30 | $1.20 | $8 |
Cost per 1,000 Requests
Short (500 tok)
$0.39
via openrouter
Medium (2K tok)
$1.56
via openrouter
Long (8K tok)
$4.80
via openrouter
Performance Estimates
Throughput by GPU
VRAM Breakdown (B200 NVL (pair), FP8)
Precision Impact
bf16
427.0 GB
weights/GPU
fp8
213.5 GB
weights/GPU
~280.0 tok/s
int4
106.8 GB
weights/GPU
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy MiniMax M3
Similar Models
MiniMax-Text-01
456B params · moe
Quality: 50
from $5.00/M
MiniMax M2.7
228.7B params · moe
Quality: 50
from $1.20/M
Llama 3.1 405B
405B params · dense
Quality: 81
from $1.00/M
Claude Opus 4.6
450B params · moe
Quality: 90
from $25.00/M
Claude Opus 4.5
400B params · moe
Quality: 90
from $25.00/M
Frequently Asked Questions
How much VRAM does MiniMax M3 need for inference?
MiniMax M3 requires approximately 795.4 GB of VRAM at BF16 precision, 397.7 GB at FP8, or 198.9 GB at INT4 quantization. Additional VRAM is needed for KV-cache (122880 bytes per token) and activations (~0.00 GB).
What is the best GPU for MiniMax M3?
The top recommended GPU for MiniMax M3 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 M3 inference cost?
MiniMax M3 API inference starts from $0.30/M input tokens and $1.20/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.