Models / Raon-1 mini Preview / VRAM
How much VRAM do you need for Raon-1 mini Preview?
Last updated:
Weight math from the model card, plus every measured RigBase result for Raon-1 mini Preview. Leave extra room for context (KV cache).
Weight size by quant
| Quant | Bits/weight | Weights only |
|---|---|---|
| Q8_0 | 8.50 | 30.6 GB |
| Q6_K | 6.56 | 23.6 GB |
| Q5_K_M | 5.70 | 20.5 GB |
| Q4_K_M | 4.89 | 17.6 GB |
| Q3_K_M | 4.00 | 14.4 GB |
| Q2_K | 3.16 | 11.4 GB |
Bits-per-weight for GGUF rows: llama.cpp quantize README. Parameter count: KRAFTON model card: 28.8B total / 4.3B activated MoE (6 of 128 routed experts + . We advise staying at Q4 or above for chat quality — see Is Q4 good enough?.
Measured speeds for Raon-1 mini Preview
| Hardware | Quant | Engine | tok/s |
|---|---|---|---|
| RTX 5070 Ti 16GB · 16 GB | IQ3_KS experts / IQ4_KS_R16 down (13.85 GB, 3.86 b | ik_llama.cpp-raon fork (CUDA 12.8) | 220 |
| RTX 5070 Ti 16GB · 16 GB | IQ3_KS experts / IQ4_KS_R16 down (13.85 GB, 3.86 b | ik_llama.cpp-raon fork (CUDA 12.8) | 190 |
| RX 7900 XTX 24GB · 24 GB | IQ4_XS GGUF | llama.cpp KRAFTON fork + PR #1 kernel pr | 75.46 |
| RX 7900 XTX 24GB · 24 GB | IQ4_XS GGUF | llama.cpp KRAFTON fork + PR #1 kernel pr | 64.57 |
| RTX 5060 Ti 16GB · 16 GB | IQ4_XS GGUF (KRAFTON, 14.97 GiB) + DFlash2 Q8_0 dr | llama.cpp KRAFTON fork fb895662 (CUDA 13 | 42.1 ⚡ |
⚡ = speculative decoding (MTP, DFlash…). Full list: best home rig for Raon-1 mini Preview.
Related
All Raon-1 mini Preview builds VRAM + KV calculator By VRAM size GGUF vs AWQ vs FP8 vs NVFP4 How much system RAM?
FAQ
How much VRAM do you need for Raon-1 mini Preview?
Weights alone are about 17.6 GB at Q4_K_M (28.8B × 4.89 bits ÷ 8). Add memory for the KV cache (context) and the runtime. The smallest-memory measured result in RigBase is 16 GB (RTX 5070 Ti 16GB) at 220 tok/s. Going below Q4 usually hurts quality for chat; see the Q4 guide.
What is the fastest measured speed for Raon-1 mini Preview?
RTX 5070 Ti 16GB: 220 tok/s plain decode (IQ3_KS experts / IQ4_KS_R16 down (13.85 GB, 3.86 bpw), q8_0 KV, ik_llama.cpp-raon fork (CUDA 12.8)).
Does system RAM replace VRAM?
Only for overflow. When the whole model fits in GPU or unified memory, system RAM barely matters for decode speed. See the system-RAM guide.
Referral link: if you sign up with Google and add $10, Runpod gives you and RigBase bonus credit (terms). It never affects our data.