Model reference · open weights
MiniCPM-o-2_6 is an open-weight language model from FriendliAI. MiniCPM-o-2_6 (BF16) weighs 17.4 GB; the smallest configuration that runs it is 2× RTX 3060 12 GB.
For self-hosting, compare this build's requirements below with the GPU you plan to rent.
Summary of the FriendliAI/MiniCPM-o-2_6 model card, 2026-10-01
What it is
| Released by | FriendliAI |
|---|---|
| Released | 2025-12-24 |
| Parameters | 8.7B |
| VRAM | 17.4 GB for the weights |
What it runs on
| Card | Requests at once | Context max | Memory | |
|---|---|---|---|---|
| 8K each | 32K each | |||
| RTX 3060 12 GB … RTX 4060 Ti 16 GB 2 smaller cards | ||||
| RTX 3090 24 GB | 8 | 2 | all 32K | 23.4 GB |
| RTX 4090 24 GB | 8 | 2 | all 32K | 23.4 GB |
| RTX 5090 32 GB | 24 | 6 | all 32K | 31.0 GB |
| L40S 48 GB | 52 | 13 | all 32K | 44.0 GB |
| A100 80 GB | 125 | 31 | all 32K | 78.2 GB |
| H100 80 GB | 115 | 28 | all 32K | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | 148 | 37 | all 32K | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | 177 | 44 | all 32K | 107 GB |
| H200 141 GB | 242 | 60 | all 32K | 138 GB |
| B200 180 GB | 322 | 80 | all 32K | 176 GB |
| 2× RTX 3060 12 GB tensor parallel | 3 | 29K | 11.6 GB a card | |
| 2× RTX 4060 Ti 16 GB tensor parallel | 19 | 4 | all 32K | 15.4 GB a card |
| Requests at once | 8K tokens each | 32K tokens each |
|---|---|---|
| 1 | 19.9 GB | 21.3 GB |
| 5 | 21.8 GB | 28.8 GB |
| 8 | 23.2 GB | 34.5 GB |
| 16 | 27.0 GB | 49.5 GB |
| 32 | 34.5 GB | 79.6 GB |
| 64 | 49.5 GB | 140 GB |
One card, with vLLM's small-card settings.