selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

RWKV logo

Can the Instinct MI50 32GB run RWKV7-G1k-13.3B-20260930?

RWKV7-G1k-13.3B-20260930·13.3B·Instinct MI50 32GB·32 GB VRAM

Saved in this browser, used on every model page.
Yes

Yes. The Instinct MI50 32GB runs RWKV7-G1k-13.3B-20260930 fully in VRAM at Q8_0 (about 22 GB of 32 GB), at roughly 40 tokens/s.

best quant
Q8_0
needs
22 GB
usable
32 GB
est. speed
~40 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the RWKV7-G1k-13.3B-20260930 page →

Every quantisation of RWKV7-G1k-13.3B-20260930 on the Instinct MI50 32GB

QuantQualityMemoryFitSpeed
Q2_K Heavy loss12 GBfits~129 tok/s
Q3_K_M Noticeable loss14 GBfits~87 tok/s
Q4_K_M Small loss16 GBfits~70 tok/s
Q5_K_M Near-lossless17 GBfits~60 tok/s
Q6_K Near-lossless19 GBfits~52 tok/s
Q8_0 pickLossless22 GBfits~40 tok/s
FP16 / BF16 Lossless34 GBoffload~11 tok/s

Speed is a bandwidth-based estimate for generating text (1024 GB/s); prompt processing and long contexts are slower.

Measured speed on the Instinct MI50 32GB

No community reports for this pair yet.

Report your own speed →

Run RWKV7-G1k-13.3B-20260930 on the Instinct MI50 32GB

Transformers · Chat with the pipeline APIdocs
from transformers import pipeline

pipe = pipeline("text-generation", model="RWKV/RWKV7-G1k-13.3B-20260930", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])

Chat models that fit the Instinct MI50 32GB comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
14.8BQ8_0≈17 GBFits in memory~36 tok/sCommercial OK
microsoft logophi-4microsoft
14.7BQ8_0≈17 GBFits in memory~36 tok/sCommercial OK
20.9BA4.2BQ8_0≈13 GBFits in memory~126 tok/sCommercial OK
8.2BQ8_0≈10 GBFits in memory~65 tok/sCommercial OK
8BQ8_0≈9.8 GBFits in memory~66 tok/sConditions
9.4BQ8_0≈11 GBFits in memory~56 tok/sCommercial OK

Other GPUs that run RWKV7-G1k-13.3B-20260930

Frequently asked questions

Can the Instinct MI50 32GB run RWKV7-G1k-13.3B-20260930?
Yes. The Instinct MI50 32GB runs RWKV7-G1k-13.3B-20260930 fully in VRAM at Q8_0 (about 22 GB of 32 GB), at roughly 40 tokens/s.
How fast does RWKV7-G1k-13.3B-20260930 run on the Instinct MI50 32GB?
An estimate from memory bandwidth gives about 40 tokens/s at Q8_0.
Which quantisation of RWKV7-G1k-13.3B-20260930 should I use on the Instinct MI50 32GB?
Q8_0: the largest one that fits. Expected quality: lossless compared with the full-precision weights.

Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.