Can Agents-A1 35B A3B run on Intel Arc Pro B60 24GB?

BARELY — Tight on Memory

A74Great

Estimated from fit model

Agents-A1 35B A3B needs ~25.9 GB VRAM. Intel Arc Pro B60 24GB has 24.0 GB. With Q4_K_M quantization, expect ~20 tok/s.

Runtime: llama.cppCapacity: OffloadBandwidth: MediumStack: StandardBottleneck: Host offload

Operating mode

Choose the run profile you care about

Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.

Current mode

Balanced

Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.

Capabilities:

Select quantization to explore

Q4_K_M (Medium quality) — 25.0 GB, 23.9 tok/s, Runs with offload (needs ~0.9 GB host RAM)

25.0 GB required24.0 GB available

104% VRAM needed

1.0 GB over capacity — needs offload or smaller quantization

Fit status

Runs with offload (needs ~0.9 GB host RAM)

Decode

23.9 tok/s

TTFT

8115 ms

Safe context

Memory

25.0 GB / 24.0 GB

Memory breakdown

Weights21.4 GB

KV Cache0.3 GB

Runtime0.9 GB

Headroom2.4 GB

See how fast it feels

See how fast it feelsAgents-A1 35B A3B on Intel Arc Pro B60 24GB
1st promptCold start — includes initialization
>Write a Python function to sort a list
Here's a Python function that sorts a list using quicksort:

def quicksort(arr):
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left   = [x for x in arr if x < pivot]
    middle = [x for x in arr if x == pivot]
    right  = [x for x in arr if x > pivot]
    return quicksort(left) + middle + quicksort(right)

# Example usage
numbers = [3, 6, 8, 10, 1, 2, 1]
print(quicksort(numbers))  # [1, 1, 2, 3, 6, 8, 10]

The algorithm picks a pivot, partitions the list into smaller, equal, and larger elements, then recursively sorts each partition and concatenates the results.
2nd promptWarm — model ready, no init overhead
>Now add type hints and a docstring
def quicksort(arr: list[int]) -> list[int]:
    """Sort a list using the quicksort algorithm.

    Args:
        arr: List of integers to sort.

    Returns:
        A new sorted list.
    """
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left   = [x for x in arr if x < pivot]
    middle = [x for x in arr if x == pivot]
    right  = [x for x in arr if x > pivot]
    return quicksort(left) + middle + quicksort(right)

Added type hints for the input and return type, plus a Google-style docstring.
Estimated: 23.9 tok/s decode · 8.1s TTFT (warm) · 60 tok/s prefill

What limits this setup

It fits through host-memory offload, and offload is the main reason performance drops.

CPU or host-memory offload is active

About 10% of the working set spills out of accelerator memory, which usually hurts latency and sustained decode throughput.

Very little memory headroom

You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.

Runtime ecosystem is narrower than CUDA

Intel GPUs can look attractive on memory per dollar, but local AI tooling, kernels, and model coverage are still broader and easier on CUDA today.

Best improvement path

Remove offload with more accelerator memory

Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.

Prefer CUDA if you want the path of least resistance

If your goal is maximum runtime coverage, easier troubleshooting, and better support for new local AI releases, CUDA is usually still the safer upgrade path.

Buy headroom, not only minimum fit

A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.

Performance by workload

Workload	Grade	Fit	Decode	TTFT	Context
Chat	A	Runs with offload	21.4 tok/s	4934 ms	4K
Coding	A	Very compromised	20.4 tok/s	9497 ms	4K
Agentic Coding	A	Very compromised	18.6 tok/s	15172 ms	4K
Reasoning	A	Very compromised	20.4 tok/s	11223 ms	4K
RAG	A	Very compromised	18.6 tok/s	18965 ms	4K

Inference speed

Agents-A1 35B A3B inference speed — tokens per second by GPU & Mac

Estimated decode speed (tokens/sec) for Agents-A1 35B A3B at Q4_K_M across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~139 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.

GPU / Mac	Memory	Quant	Speed (tok/s)	Fits?
RTX 5090 32GB	32 GB	Q4_K_M	139.1	Fits
RTX 4090 24GB	24 GB	Q4_K_M	77.7	Offloads
Mac Studio M3 Ultra 256GB	256 GB	Q4_K_M	76.8	Fits
RX 7900 XTX 24GB	24 GB	Q4_K_M	65.5	Offloads
Mac Studio M2 Ultra 128GB	128 GB	Q4_K_M	64.0	Fits
RTX 3090 24GB	24 GB	Q4_K_M	62.1	Offloads
Mac Studio M1 Ultra 128GB	128 GB	Q4_K_M	60.7	Fits
MacBook Pro M4 Max 128GB	128 GB	Q4_K_M	47.4	Fits
MacBook Pro M4 Max 64GB	64 GB	Q4_K_M	47.4	Fits
MacBook Pro M3 Max 64GB	64 GB	Q4_K_M	36.4	Fits
MacBook Pro M1 Max 64GB	64 GB	Q4_K_M	33.4	Fits
MacBook Pro M4 Pro 48GB	48 GB	Q4_K_M	31.9	Fits
RTX 4080 Super 16GB	16 GB	Q4_K_M	28.3	Too big
RTX 4070 12GB	12 GB	Q4_K_M	9.9	Too big
RTX 3060 12GB	12 GB	Q4_K_M	5.8	Too big
RTX 4060 8GB	8 GB	Q4_K_M	4.4	Too big

Estimates for single-stream decoding at Q4_K_M; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.

Quantization options

How Agents-A1 35B A3B (35.099998474121094B params) fits at each quantization level on Intel Arc Pro B60 24GB (24.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q1_0_G128	1.125	5.1 GB	Very Low	A81
Q2_0_G128	1.71	9.4 GB	Low	A84
Q2_K	2	13.7 GB	Low	A84
Q3_K_SBest for your GPU	3	17.2 GB	Low	A84
NVFP4	4	19.7 GB	Medium	F0
Q4_K_M	4	21.4 GB	Medium	F0
Q5_K_M	5	25.3 GB	High	F0
Q6_K	6	28.8 GB	High	F0
Q8_0	8	37.6 GB	Very High	F0
F16	16	72.0 GB	Maximum	F0

Get started

Copy-paste commands to run Agents-A1 35B A3B on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "InternScience/Agents-A1" \
  --hf-file "Agents-A1-Q4_K_M.gguf" \
  -c 4096 -ngl 99

Frequently asked questions

Can Intel Arc Pro B60 24GB run Agents-A1 35B A3B?

Yes, Intel Arc Pro B60 24GB can run Agents-A1 35B A3B with a A grade (Very compromised). Expected decode speed: 20.4 tok/s.

How much VRAM does Agents-A1 35B A3B need?

Agents-A1 35B A3B (35.099998474121094B parameters) requires approximately 25.9 GB of memory with Q4_K_M quantization.

What is the best quantization for Agents-A1 35B A3B?

The recommended quantization for Agents-A1 35B A3B is Q4_K_M, which balances quality and memory efficiency.

What speed will Agents-A1 35B A3B run at on Intel Arc Pro B60 24GB?

On Intel Arc Pro B60 24GB, Agents-A1 35B A3B achieves approximately 20.4 tokens per second decode speed with a time-to-first-token of 9497ms using Q4_K_M quantization.

Can Intel Arc Pro B60 24GB run Agents-A1 35B A3B for coding?

For coding workloads, Agents-A1 35B A3B on Intel Arc Pro B60 24GB receives a A grade with 20.4 tok/s and 4K context.

What context window can Agents-A1 35B A3B use on Intel Arc Pro B60 24GB?

On Intel Arc Pro B60 24GB, Agents-A1 35B A3B can safely use up to 4K tokens of context. The model's official context limit is 262K, but available memory constrains the safe maximum.

What should I upgrade first if Agents-A1 35B A3B feels slow on Intel Arc Pro B60 24GB?

Remove offload with more accelerator memory. Prioritize a GPU or unified-memory tier that fits the whole model natively. Removing offload usually helps more than small compute gains.

Would CUDA be a better path than Intel Arc Pro B60 24GB for Agents-A1 35B A3B?

Often yes, if your goal is the easiest setup and the widest runtime support. Intel can offer attractive memory capacity, but CUDA still tends to win on tooling maturity, guides, kernels, and model coverage for local AI.

See all results for Intel Arc Pro B60 24GB See all hardware for Agents-A1 35B A3B

Embed this result▼

Paste this snippet into any page to show a live fit card.

<iframe src="https://willitrunai.com/embed/agents-a1-on-arc-pro-b60-24gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>

Preview: