Your sovereign, private Minimax 3
MiniMax-M3 is a native multimodal model with ~428B total and ~23B activated parameters and a 1M-token context. Its MiniMax Sparse Attention (MSA) delivers 9x prefill and 15x decode speedups over M2 at 1M context, and it achieves frontier-level performance on long-horizon agentic coding and cowork benchmarks with adaptive reasoning modes. A reasoning model in the medium tier, served on ToothFairyAI's AU, EU and US endpoints — private by default, billed in US dollars per million tokens at the published rates below. No training on your data, no vendor lock-in.
How Minimax 3 compares
Independent intelligence index for Minimax 3 against every peer in the Medium tier (and the tier above where needed) — same source, same tests.
Latency — end-to-end seconds for a 500-token answer
Capability indexes
Domain-weighted agentic performance for Minimax 3.
Capability index vs medium peers
The evaluations behind the index
Independent benchmark results for Minimax 3 — Elo scores as published; pass rates shown as percentages.
Per-evaluation economics
| Evaluation | Score | Time / task | Output tokens / task |
|---|---|---|---|
| AA-Briefcase | 1091.10 | 11.4 min | 81897 |
| GDPval-AA | 1230.07 | 6.0 min | 43087 |
| AutomationBench-AA | 21.3% | 2.9 min | 20791 |
| Terminal-Bench 4.0 | 2.0% | 24.6 min | 176730 |
| SciCode | 47.1% | 2.3 min | 16494 |
| Humanity's Last Exam | 39.0% | 3.0 min | 21485 |
| GDP.pdf | 9.8% | 1.1 min | 7937 |
| CritPt | 3.7% | 10.9 min | 77915 |
| AA-Omniscience | 1.35 | 0.4 min | 2883 |
| AA-LCR (long context) | 83.0% | 0.2 min | 1409 |
Run Minimax 3 privately
One key, one endpoint, three jurisdictions.
- Size tierMedium
- Serving routeStandard
- Input rate$0.42/1M
- Output rate$1.51/1M
- Cached input rate$0.08/1M
- Context window180K tokens
- Max output16K tokens
- ReasoningYes — emits thinking traces
- Vision & video inputSupports image input
- Tool callingSupported
- Data residencyAU, EU and US endpoints — processing stays inside the region you call