MiMo-V2.5-Pro

Pricing — 1 offering(s)

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

Capability profile

Operator guidance

Route here for token-efficient agentic and coding workloads at very long context and low cost — its stated strengths. Benchmark signal is strong but comes largely from independent coverage rather than a Xiaomi-published technical report, so treat the "near Opus 4.6 / GPT-5.4" framing as directional. Against the other cheap 1M-context MoE flagships in this registry (DeepSeek V4-Flash, GLM-5.2), MiMo-V2.5-Pro's distinguishing claim is trajectory token-efficiency.

Use cases

• Long-horizon agentic workflows with many sequential tool calls, where token efficiency per trajectory matters to cost
• Agentic coding tasks at open-API cost with a 1M context window
• High-volume production use via an OpenAI/Anthropic-compatible first-party API

Limitations

• No Xiaomi-published technical report located; architecture and benchmark figures are from independent coverage and Artificial Analysis
• Reasoning / instruction-following / multilingual have no standalone published benchmark — rated conservatively or `moderate`
• Open weights: XiaomiMiMo/MiMo-V2.5 on Hugging Face, but license terms not independently confirmed in this pass
• Registry pricing has a later CNY-converted `confidence: estimated` price-monitor proposal layered on top of the first-party figure

Citations

• MiMo-V2.5-Pro — Artificial Analysis (independent intelligence/price analysis)
• XiaomiMiMo/MiMo-V2.5 (Hugging Face)
• MiMo-V2.5-Pro — Xiaomi
• MiMo pay-as-you-go pricing (Xiaomi docs)