Arc Pro B70

Muse Glimmer 30B is the best use of this card: 18.7 GB of the 32 GB available, an estimated 25 tokens per second, usable, a little slow, and 131,072 tokens of context still available.

How this is chosen, since no benchmark is quoted: models are compared by size, a mixture-of-experts model counting at the geometric mean of its total and active parameters (a rule of thumb, not a measurement). The pick is the biggest class that fits with memory to spare and answers at 25 tokens per second or better, and within that class the most recent general-purpose release; coding and reasoning specialists are listed above but not recommended by default.

Qwen3.6 35B-A3B (MoE) is the largest model the card will hold, at 22.4 GB, with room for 131,072 tokens of context. It is a mixture-of-experts model reading only 3B parameters per token, so it is the faster of the two at an estimated 245 tokens per second; what it gives up is the depth of a dense model that reads all of its weights for every token.

If quality matters more than parameter count, Mistral Small 3.1 24B fits at Q8_0 in about 28.3 GB. Q8 is near-lossless; a smaller model at Q8 often beats a larger one squeezed into Q3.

Quantising the KV cache to Q8 roughly doubles what you can hold, at a quality cost most people never notice: Gemma 4 31B on this card goes from 123,904 to 131,072 tokens. Capped at 128k tokens, or at the model's own native window where that is smaller, marked "model limit": free memory beyond that point buys nothing. Rows that hold far more context than their size suggests are hybrid, sliding-window or latent-attention models, which cache only a fraction of what a standard transformer does; each model page shows the working.

The first model out of reach is Llama 3.3 70B: about 45.8 GB at Q4_K_M and 8k context, against 32 GB of usable memory. Dropping to Q3_K_M would need about 37.7 GB, which still does not fit. Offloading the remainder to system RAM works and is ten to fifty times slower; it is a way to see a model run, not a way to use one.

Worth it when you need 32 GB on one card at modest power and your tools run on Intel, typically llama.cpp, or vLLM under LLM Scaler on Linux. If you need speed, an RTX 5090 reads 32 GB nearly three times as fast and runs CUDA, at 575 W. For occasional 32 GB work, renting by the hour costs less than owning.

Workstation cards are sold on quotes rather than shelf prices, so there is no single number to publish. Put the figure you are actually looking at, with your electricity rate and your real monthly hours, into the build vs rent calculator, which already knows this card draws 230 W under load. It returns the month owning becomes cheaper, or tells you it never does. Hours per month decides it far more often than the price does.

• Radeon AI PRO R9700: 32 GB · 640 GB/s · best fit Muse Glimmer 30B at ~26 tok/s. AMD's 32 GB workstation card: RX 9070 XT bandwidth with twice the memory, at 300 W.
• Arc Pro B60: 24 GB · 456 GB/s · best fit Gemma 4 12B at ~46 tok/s. Intel's 24 GB workstation card: the B580's bandwidth with twice the memory.
• RTX 5090: 32 GB · 1,792 GB/s · best fit Muse Glimmer 30B at ~73 tok/s. The first consumer card whose memory bandwidth changes what is comfortable.

All 39 cards are compared side by side on the GPU index, and the same numbers from the model's point of view are on each model fit table. To check one specific pairing rather than read a table, use the can-I-run-it checker.