Can the Arc B570 run Codestral 22B v0.1?
Only with CPU offload at Q4_K_M.
Codestral 22B v0.1 is 22.2B parameters. At Q4_K_M the weights are about 12.4 GiB; with the 1.5 GiB runtime reserve that is 13.9 GiB against the card's 10GB, about 3.9 GiB more than the card has.
All three quants
10GB card| Quant | Weights | With reserve | Verdict | Headroom |
|---|---|---|---|---|
| Q4_K_Mthe everyday quant | 12.4 GiB | 13.9 GiB | Only with CPU offload | short 3.9 GiB |
| Q8_0near-lossless | 21.9 GiB | 23.4 GiB | Does not fit | short 13.4 GiB |
| FP16full weights | 41.4 GiB | 42.9 GiB | Does not fit | short 32.9 GiB |
Verdict rules: fits comfortably when weights plus reserve sit within 85% of VRAM; tight when they fit with little left for context; CPU offload when they exceed VRAM by up to 2x (runs, slowly, with layers in system RAM); no beyond that. The KV cache per token depends on the model's architecture and is not modelled here.
The numbers for this pairing
At Q4_K_M the weights take about 12.4 GiB, so with the 1.5 GiB runtime reserve it needs 13.9 GB: 3.9 GB more than the card's 10 GB.
No version of it fits entirely in this card's memory, so it would lean on system RAM and run far slower.
The cheapest card on our list that runs it comfortably at Q4 is the Radeon RX 7900 XTX at about $850, against about $260 for this card.
At about $260, this card costs roughly $26 per GB of VRAM.
Running it at its full 150 W board power for 4 hours a day would use about 18 kWh a month, about $3.31 at the US average residential rate of 18.4 cents per kWh (2026-05); real use sits below that ceiling.
Prices are the lowest current listings our tracker saw on Oct 7, 2026, since averages include premium variants; launch prices are marked where we have no listing.
Not this card. These fit it comfortably at Q4.
- GeForce RTX 5090 32GB · headroom 18.1 GiB
- GeForce RTX 3090 (used) 24GB · headroom 10.1 GiB
- GeForce RTX 4090 (used) 24GB · headroom 10.1 GiB
- Radeon RX 7900 XTX 24GB · headroom 10.1 GiB
What the Arc B570 does run well → · Codestral 22B v0.1 on every card →
About the model
Codestral 22B v0.1 by Mistral AI: 32,768-token native context, text, released 2024 under the Mistral AI Non-Production License. Model card →