Four Tiers, Same Models, Different Fuel Tank
MiMo Token Plan's four tiers differ only in Credits volume. From Lite at ¥39/month to Max at ¥659/month, you get full access to all 6 V2.5 models, support for OpenCode, Claude Code, Cline, MiMo Code, and every other tool, plus the same nighttime 0.8x discount and free TTS. What you pay determines only how many Credits you can spend each month (source).
| Tier | Monthly | Annual (per month) | Credits/Month | Max Pro model output tokens* |
|---|---|---|---|---|
| Lite | ¥39 | ¥34.32 | 4.1B | ~6.8M tokens |
| Standard | ¥99 | ¥87.12 | 11B | ~18.3M tokens |
| Pro | ¥329 | ¥289.52 | 38B | ~63.3M tokens |
| Max | ¥659 | ¥579.92 | 82B | ~136.7M tokens |
* Estimated using Pro model output rate of 600 Credits per 1M tokens. Actual consumption varies by model choice, cache hit rate, and context length.
How Credits Are Deducted
Three Models, Three Burn Rates
Not all models consume Credits at the same pace. The Pro model (mimo-v2.5-pro) burns through Credits fastest, the Standard model (mimo-v2.5) sits in the middle, and TTS/ASR follow separate rules (source):
| Model | Cache Hit | Uncached Input | Output | Unit |
|---|---|---|---|---|
| MiMo-V2.5-Pro | 2.5 | 300 | 600 | Credits / 1M tokens |
| MiMo-V2.5 (Standard) | 2 | 100 | 200 | Credits / 1M tokens |
| ASR | — | 30M/hr | 30M/hr | Credits / audio hour |
| TTS Series (3 models) | Free | Free | Free | No Credit charge |
The gap between cache hits and uncached input is massive—for the Pro model at 1M tokens, a cache hit costs 2.5 Credits while uncached costs 300, a 120x difference. Workflows that reliably leverage context caching (e.g., lengthy multi-turn sessions on the same project) can substantially slow down Credit consumption.
Real-World Budgeting: 4 Workload Scenarios
Using the Pro tier (38B Credits/month) as baseline:
Scenario A: 3–4 hours daily moderate coding, Pro model, ~40% cache hit rate, ~120M Credits/day. Monthly: ~3.6B → Standard is tight, Pro is comfortable.
Scenario B: Full-day intensive Agent workflows, Pro model, 300–500M Credits/day. Monthly: 10–15B → Pro is insufficient, you need Max.
Scenario C: Light daily use, mostly Standard model, 100–150M Credits/day. Monthly: 3.5–4.5B → Lite barely fits, Standard is safe.
Scenario D: Nighttime-only usage, all operations at 0.8x. Pro tier's 38B Credits behave like 47.5B in daytime purchasing power → nighttime boosts effective value by 25%.
Stacking Discounts: 12% Off + Night ×0.8
MiMo discounts operate on two layers: how much you pay for the subscription, and how fast Credits drain.
Purchase discounts:
- First month 12% off: one-time, renewal returns to full price
- Annual billing 12% off: auto-debited monthly, discount applies year-round
- Referral 10% off first purchase: both parties get ¥10 credits
Usage discounts:
- 0:00–8:00 Beijing time, Credits ×0.8 = 25% more effective quota
Best combination: Annual Max + nighttime usage. Annual Max at ¥579.92/month with nighttime discount effectively delivers the purchasing power of 1,025B Credits at daytime rates—cheaper than stacking two daytime Pro subscriptions.
Which Tier Fits You
Lite (¥39/month): trial and light use. First month ¥34.32 is barely a coffee, great for a test run in Claude Code. Not recommended for daily heavy use—4.1B Credits at Pro model rates works out to roughly 1,300 moderate sessions, which a frequent coder can burn through in a week.
Standard (¥99/month): everyday coding. 11B Credits with the Standard model (200 Credits/1M output) comfortably covers daily workflows. Pro tip: pair Standard-tier model + Standard plan rather than Pro model + Standard plan—the latter burns Credits too fast, and going all-in on Pro plan with Pro model is actually more cost-effective at that point.
Pro (¥329/month): professional core developers. 38B Credits lets you use Pro model without watching the meter—long-chain agent tasks, large-scale refactoring, parallel projects. This is the sweet spot for most paying users.
Max (¥659/month): team or heavy agent users. 82B Credits targets all-day high-concurrency usage. Ideal for team sharing or treating MiMo as your primary coding foundation. With annual billing + nighttime scheduling, cost-efficiency can beat Pro.
Caveats
-
V2 series retired June 30, 2026. If you see
mimo-v2-proormimo-v2-omniin old docs, those model names are dead. Switch directly to V2.5 equivalents (source). -
Upgrade only, no downgrade mid-cycle. You can pay the difference to go from Lite to Standard, but dropping from Max to Pro requires waiting for the current cycle to end.
-
API Keys come in two types. Token Plan uses
tp-xxxxx-formatted keys with separate Base URLs. Don't use Token Plan endpoints with open API keys or vice versa. -
Free TTS isn't permanent. VoiceClone, VoiceDesign, and base TTS currently cost zero Credits, but Xiaomi may announce pricing any time. Budget accordingly if your product depends on speech output.
-
Model capability is identical across all four tiers. Choosing Lite does not lock you out of Pro model—mimo-v2.5-pro is available on every plan, it just burns Credits faster on lower tiers.