Can the RTX 5090 run Axion-1.0-35B?
Axion-1.0-35B·35.1B (3.9B active)·GeForce RTX 5090·32 GB VRAM
Yes. The RTX 5090 runs Axion-1.0-35B fully in VRAM at Q6_K (about 29 GB of 32 GB), at roughly 307 tokens/s.
- best quant
- Q6_K
- needs
- 29 GB
- usable
- 32 GB
- est. speed
- ~307 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Axion-1.0-35B page →
Every quantisation of Axion-1.0-35B on the RTX 5090
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 14 GB | fits | ~768 tok/s |
| Q3_K_M | 18 GB | fits | ~517 tok/s |
| Q4_K_M | 22 GB | fits | ~417 tok/s |
| Q5_K_M | 25 GB | fits | ~355 tok/s |
| Q6_K pick | 29 GB | fits | ~307 tok/s |
| Q8_0 | 37 GB | offload | ~47 tok/s |
| FP16 / BF16 | 70 GB | too big | — |
Speed is a bandwidth-based estimate for generating text (1792 GB/s); prompt processing and long contexts are slower.
Measured speed on the RTX 5090
No community reports for this pair yet.
Report your own speed →Run Axion-1.0-35B on the RTX 5090
Using Q6_K, the best fit for RTX 5090.
ollama run hf.co/mradermacher/Axion-1.0-35B-GGUF:Q6_K llama-server -hf mradermacher/Axion-1.0-35B-GGUF:Q6_KDownloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.
lms get mradermacher/Axion-1.0-35B-GGUF@Q6_KThen load it from the LM Studio app or with `lms load`.
vllm serve Axion1-0/Axion-1.0-35Bservices:
vllm:
image: vllm/vllm-openai:latest
command: Axion1-0/Axion-1.0-35B
ports:
- "8000:8000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 8000.
sglang serve Axion1-0/Axion-1.0-35B --host 0.0.0.0 --port 30000services:
sglang:
image: lmsysorg/sglang:latest
command: sglang serve Axion1-0/Axion-1.0-35B --host 0.0.0.0 --port 30000
ports:
- "30000:30000"
ipc: host
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]API on port 30000.
from transformers import pipeline
pipe = pipeline("text-generation", model="Axion1-0/Axion-1.0-35B", 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 RTX 5090 comfortably
All that fit →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 36BA3B | Q5_K_M | ≈26 GBFits in memory | ~462 tok/s | |
| 31.2BA4.6B | Q5_K_M | ≈29 GBFits in memory | ~301 tok/s | |
| 32.8B | Q6_K | ≈29 GBFits in memory | ~37 tok/s | |
| 39.8BA1.3B | Q5_K_M | ≈29 GBFits in memory | ~1108 tok/s | |
| 20.9BA4.2B | Q8_0 | ≈13 GBFits in memory | ~221 tok/s | |
| 30.5BA3B | Q6_K | ≈26 GBFits in memory | ~399 tok/s |
Other GPUs that run Axion-1.0-35B
Frequently asked questions
- Can the RTX 5090 run Axion-1.0-35B?
- Yes. The RTX 5090 runs Axion-1.0-35B fully in VRAM at Q6_K (about 29 GB of 32 GB), at roughly 307 tokens/s.
- How fast does Axion-1.0-35B run on the RTX 5090?
- An estimate from memory bandwidth gives about 307 tokens/s at Q6_K.
- Which quantisation of Axion-1.0-35B should I use on the RTX 5090?
- Q6_K: the largest one that fits. Expected quality: near-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.