AI Index

MiniMax-M3

Language MiniMax Verified Released Jun 1, 2026
  • In TextImageVideo Out Text
  • Reasoning
  • Reasoning can be off
  • Tool use
  • Open weights

Prices

2 offerings · prices as listed
MiniMax API Official Global · USD MiniMax-M3
Conditions/ 1M tokensInputCache readOutput
Standard $0.30$0.060$1.20
Prompts over 512K tokens $0.60$0.12$2.40
Priority tier $0.45$0.090$1.80
Prompts over 512K tokensPriority tier $0.90$0.18$3.60

Prices from Pricing Overview - MiniMax API Docs (checked Nov 1, 2026)

MiniMax API Official China · CNY MiniMax-M3
Conditions/ 1M tokensInputCache readOutput
Standard ¥2.10¥0.42¥8.40
Prompts over 512K tokens ¥4.20¥0.84¥16.80
Priority tier ¥3.15¥0.63¥12.60
Prompts over 512K tokensPriority tier ¥6.30¥1.26¥25.20

Prices from MiniMax 按量计费价格 (checked Nov 1, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Family
minimax-m
Model ID

Not stated on MiniMax’s pages: max input, reasoning efforts, max thinking budget, structured output, built-in tools, knowledge cutoff, licence, aliases.

Run locally

Weights: MiniMaxAI/MiniMax-M3
  • Parameters 427B
  • Active 23B · 128 experts, 4 per token
  • Native precision BF16
  • Layers 60
  • KV cache per token 120 KB
  • Config positions 1024K (config.json max_position_embeddings)
  • API context 977K (maker’s API)
My hardware GeForce RTX 4090 24 GB · 1008 GB/s · 64 GB DDR5 dual channel · 89.6 GB/s
On your hardware · GeForce RTX 4090
Doesn’t fit IQ4_XS (automatic) 8K context
WeightsKV cacheOverheadUsable
Needs
≈ 222.8 GB (weights 211.3 + KV cache 0.9 + overhead 10.6)
Usable
21.6 GB (90% of its memory)
Try
it’s too large for this hardware; see the smallest that runs it below
PrecisionWeightsNeeds at 8KOn your hardwareLongest context
BF16 (native)maker’s file795.5 GB≈ 836.2Doesn’t fit—
Q8_0422.6 GB≈ 444.6Doesn’t fit—
FP8397.7 GB≈ 418.5Doesn’t fit—
Q6_K326.6 GB≈ 343.9Doesn’t fit—
Q5_K_M282.9 GB≈ 298.0Doesn’t fit—
Q5_0273.4 GB≈ 288.0Doesn’t fit—
Q4_K_M241.6 GB≈ 254.6Doesn’t fit—
Q4_0223.7 GB≈ 235.8Doesn’t fit—
IQ4_XS211.3 GB≈ 222.8Doesn’t fit—
Q3_K_Mquality drops194.4 GB≈ 205.0Doesn’t fit—
IQ3_XXSquality drops152.1 GB≈ 160.7Doesn’t fit—
Q2_Kquality drops149.1 GB≈ 157.5Doesn’t fit—
IQ2_XXSquality drops102.4 GB≈ 108.5Doesn’t fit—

Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.

Compared with other hardware

Each at the highest precision it runs. Click one for details.

Smallest that runs it (Q4_K_M, 8K): no single listed GPU; unified memory Apple M5 Ultra (80-core GPU) · 512 GB.

Run commands

The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.

Memory to fine-tune
Full≈ 6367.2 GBDoesn’t fit your hardware
LoRA≈ 799.9 GBDoesn’t fit your hardware
QLoRA≈ 228.1 GBDoesn’t fit your hardware

Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.

Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated

Cited scores

From third-party leaderboards, not measured by us
  • Epoch Capabilities Index 147.6

    #67 of 118 listed models · CI 145.4–149.8

  • LMArena Text 1,439

    #64 of 163 listed models · CI 1,436–1,443 · 68,611 votes

  • Epoch AI · Chess Puzzles 17%

    #69 of 85 listed models · ±3.78

    Also listed: reasoning off 8%

  • Epoch AI · GPQA Diamond 90.9%

    #32 of 168 listed models · ±2.05

    Also listed: reasoning off 81.3%

  • Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Oct 28 – Nov 5, 2026
  • LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Sep 13 – Oct 28, 2026
  • EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 24, 2026
  • LiveBench (White et al.), release 2026-06-25, livebench.ai, licensed CC BY-SA 4.0. This board, adapted from it, is shared under CC BY-SA 4.0. · License · as of Jun 25, 2026
  • changes we made

Included in coding plans

All coding plans →

Changes

All changes →

No changes recorded yet.

Sources

MiniMax’s official policies: data use, privacy, terms →

MiniMax · China · minimax.io

Snapshots, the same family and other models from this provider