Qwen 3 235B.
Alibaba's frontier MoE — 235B total / 22B active.
1× AMD MI300.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
Smaller models distilled from Qwen 3 235B.
Lightweight student models trained to mimic Qwen 3 235B's outputs.
Variants in the Qwen 3 family.
Frequently asked.
How do I run Qwen 3 235B?
Where can I access Qwen 3 235B?
How much does it cost to run Qwen 3 235B?
Is Qwen 3 235B open-source or proprietary?
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP16
FP16 — half precision (default)
|
768 GB | — | — | Compare → | |
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
384 GB | — | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
192 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Qwen 3 235B across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
No priced API access rows on file for Qwen 3 235B yet.
Rent the GPU instead of paying per token.
For an open-weights model like Qwen 3 235B, you can rent a GPU and serve inference yourself. The math: cheapest GPU rental × 730 hours/month + your electricity rate × power draw.
Assumes the GPU runs 24/7 at ~85% utilisation. If your traffic is bursty, you'll pay less for the API and probably more for the GPU (idle hours still cost rental). The breakeven analysis lives on the Self-host vs API breakeven tool.
About Qwen 3 235B.
Qwen 3 235B is Alibaba's flagship MoE released April 2025. Apache-2.0 licensed (a significant license upgrade from the 2.5 family). 235B total, 22B active per token, 128K context. Notable thinking mode that can be toggled per request.