NVIDIA96 GB VRAM1,792 GB/s$1.5/hr

GPT-OSS 20B (MoE) on RTX PRO 6000 Blackwell 96GB: 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 RTX PRO 6000 Blackwell 96GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
22.6 GB needed / 96 GB available
Decode speed
485.3
tok/s per stream
Aggregate
5824
tok/s @ 8 concurrent
Cost / 1M tokens
$0.0715
1× RTX PRO 6000 Blackwell 96GB
What changes as you add RTX PRO 6000 Blackwell 96GBs
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× RTX96 GB22.6 GB485.35824975.8 ms$0.0715✓
2× RTX192 GB22.6 GB825.119802503.1 ms$0.0421✓
4× RTX384 GB22.6 GB165079206266.5 ms$0.0210✓
8× RTX768 GB22.6 GB3300316826148.3 ms$0.0105✓
Concurrency vs context length
On 1× RTX PRO 6000 Blackwell 96GB 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?
4K723206.2 MB✓ fits
8K367412.3 MB✓ fits
33K921.61 GB✓ fits
66K463.22 GB✓ fits
131K236.44 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 RTX PRO 6000 Blackwell 96GB?

Yes. GPT-OSS 20B (MoE) at FP8 (E4M3) needs 22.6 GB including the KV cache, and RTX PRO 6000 Blackwell 96GB has 96 GB — so it fits on a single GPU.

How fast is GPT-OSS 20B (MoE) on RTX PRO 6000 Blackwell 96GB?

485.3 tokens/sec per stream and 5824 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 975.8 ms, and inter-token latency is 2.1 ms.

What does GPT-OSS 20B (MoE) on RTX PRO 6000 Blackwell 96GB cost per million tokens?

$0.0715 per 1M output tokens at 1× RTX PRO 6000 Blackwell 96GB ($1.50/hr). 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 RTX PRO 6000 Blackwell 96GB?

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 RTX PRO 6000 Blackwell 96GB the right choice for GPT-OSS 20B (MoE)?

It works. Compare $0.0715 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