Apple64 GB VRAM400 GB/s$0/hr

GPT-OSS 20B (MoE) on Mac M3 Max (64GB): Does It Fit, How Fast, What It Costs

GPT-OSS 20B (MoE) at FP8 (E4M3) needs 20.9 GB for its weights and 22.6 GB in total at 8K context with 8 concurrent users. That means it fits on a single Mac M3 Max (64GB). Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
22.6 GB needed / 64 GB available
Decode speed
108.3
tok/s per stream
Aggregate
972.2
tok/s @ 8 concurrent
Cost / 1M tokens
—
1× Mac M3 Max (64GB) · quote required
What changes as you add Mac M3 Max (64GB)s
Tensor-parallel degree vs VRAM, throughput, and cost. Cost scales linearly with GPU count; decode throughput scales near-linearly too, so cost per token stays roughly flat.
GPUsTotal VRAMVRAM neededDecode tok/sAggregate tok/sTTFTCost / 1M tokFits?
1× Mac64 GB22.6 GB108.3972.28.47 s—✓
2× Mac128 GB22.6 GB184.238894.25 s—✓
4× Mac256 GB22.6 GB368.3155562.14 s—✓
8× Mac512 GB22.6 GB736.7622221.08 s—✓
Concurrency vs context length
On 1× Mac M3 Max (64GB) at FP8 (E4M3). KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K415206.2 MB✓ fits
8K210412.3 MB✓ fits
33K531.61 GB✓ fits
66K263.22 GB✓ fits
131K136.44 GB✓ fits
Other GPUs that run GPT-OSS 20B (MoE)
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB1×5824$0.0238
RTX 3090 24GB1×2472$0.0281
RTX 4090 24GB1×3276$0.0339
AMD MI300X 192GB1×17225$0.0403
V100 SXM2 32GB1×2925$0.0475
A100 40GB SXM41×5038$0.0496
Frequently asked questions

Can GPT-OSS 20B (MoE) run on Mac M3 Max (64GB)?

Yes. GPT-OSS 20B (MoE) at FP8 (E4M3) needs 22.6 GB including the KV cache, and Mac M3 Max (64GB) has 64 GB — so it fits on a single GPU.

How fast is GPT-OSS 20B (MoE) on Mac M3 Max (64GB)?

108.3 tokens/sec per stream and 972.2 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 8.47 s, and inter-token latency is 9.2 ms.

What does GPT-OSS 20B (MoE) on Mac M3 Max (64GB) cost per million tokens?

— per 1M output tokens at 1× Mac M3 Max (64GB). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does GPT-OSS 20B (MoE) need on Mac M3 Max (64GB)?

20.9 GB for the weights at FP8 (E4M3), plus 1.6 GB of KV cache for 8 concurrent requests at 8K context — 22.6 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is Mac M3 Max (64GB) the right choice for GPT-OSS 20B (MoE)?

It works. Compare — per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.006615, and the fastest is 1× Cerebras CS-3 (WSE-3) at 65000 tok/s.

Adjust the workload yourself

Same formulas, your context length, concurrency, and quantisation.

Open in calculator