Xiaomi MiMo Token Plan Max logo

Xiaomi MiMo Token Plan Max

LLM 3.2/10

82B credits/mo (the famous '82 billion' — credits, not tokens): ~4B real tokens/mo on mimo-v2.5-pro or ~11B on v2.5 in cache-heavy coding (~12,800 medium/complex task rounds); same OpenAI + Anthropic endpoints; 0.8x night discount

Xiaomi MiMo Token Plan Max scores 3.2/10 for free value in LLM. The main constraint: Paid — $100/mo (¥659; annual ~$88/mo). 82B credits/mo (the famous '82 billion' — credits, not tokens): ~4B real tokens/mo on mimo-v2.5-pro or ~11B on v2.5 in cache-heavy coding (~12,800 medium/complex task rounds); same OpenAI + Anthropic endpoints; 0.8x night discount.

Category LLM Kind Paid plan Caveats CREDITS are NOT tokens (per-token credit deduction varies by cache hit); Token Plan ToS restricts use to coding tools — automated scripts and custom app backends are PROHIBITED (ban risk; use MiMo pay-as-you-go API for agents); no refunds; credits reset monthly; Chinese-model content censorship applies

Free value score

3.2 / 10
Show the math
Model Quality 7.6/10
value = 42 artificial_analysis_score
weighted score = 0.2 * 7.6/10
Access 10/10
value = 10/10
weighted score = 0.1 * 10/10
Price Efficiency 1/10
value = 0.01 value/$
weighted score = 0.7 * 1/10

Key metrics

MiMo-V2.5-Pro 42 Best model
0.01x Value ratio

What you get

Paid — $100/mo (¥659; annual ~$88/mo). 82B credits/mo (the famous '82 billion' — credits, not tokens): ~4B real tokens/mo on mimo-v2.5-pro or ~11B on v2.5 in cache-heavy coding (~12,800 medium/complex task rounds); same OpenAI + Anthropic endpoints; 0.8x night discount

~4B Pro / ~11B v2.5 effective tokens/mo for $100 => up to ~110M tokens per $1 at massive scale; MiMo-V2.5-Pro is omnimodal (vision+audio) with 1M context

Gotchas

No gotchas documented yet. Contribute one if you know a constraint we missed.

Sources

paid value llm-providers
Open service ← All LLM

Alternatives

Same category · similar quota
airouter.ch Flat Rateairouter.ch Flat Rate · paid · Paid plan
3.6

'Unlimited API requests, no token counting' (fair-use: personal/dev/testing) for Qwen3.6 (262K ctx, reasoning, vision) + DeepSeek-V4 + DeepSeek-V4-Flash; OpenAI-compatible endpoint, Swiss-hosted, markets itself OpenClaw-friendly

0.02×39.6 AA0.0×
May be outdatedChecked 23d ago
NoncomPersonalPaid only
Chutes PlusChutes Plus · paid · Paid plan
4.0

Serverless open-model inference (decentralized GPU); Plus tier monthly subscription with daily request quota across open-source models

0.1×24.7 AA0.1×
May be outdatedChecked 26d ago
Flat ratePaid only
NanoGPT SubscriptionNanoGPT Subscription · paid · Paid plan
5.1

GLM 5.2 + Kimi K2.7 Code High-Speed (~180-260 tok/s) + 200+ open models; 60M input tokens/week (GLM/Kimi count 2x => ~30M effective), 10 concurrent; OpenAI- AND Anthropic-compatible API + no-markup pay-as-you-go

0.13×51.1 AA0.1×
May be outdatedChecked 23d ago
Paid onlyEmbeddings