NVIDIA48 GB VRAM864 GB/s$0.7/hr

Mistral Large 3 (256K) on L40S 48GB: Does It Fit, How Fast, What It Costs

Mistral Large 3 (256K) at AWQ 4-bit needs 129.3 GB for its weights and 150.2 GB in total at 8K context with 8 concurrent users. That means it needs 4× L40S 48GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 4×
150.2 GB needed / 192 GB available
Decode speed
43.03
tok/s per stream
Aggregate
2065
tok/s @ 8 concurrent
Cost / 1M tokens
$0.3766
4× L40S 48GB

L40S 48GB has no NVLink, so 4-way tensor parallelism runs over PCIe. tokcalc applies a flat 0.85× multi-GPU efficiency factor for this, which is realistic for NVLink-equipped datacenter cards but optimistic for PCIe scale-out — treat the throughput and cost figures for this configuration as an upper bound, and measure before committing.

What changes as you add L40S 48GBs
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× L40S48 GB150.2 GB0.0000.000——✗
2× L40S96 GB150.2 GB0.0000.000——✗
4× L40S192 GB150.2 GB43.0320656.40 s$0.3766✓
8× L40S384 GB150.2 GB86.0682623.22 s$0.1883✓
Concurrency vs context length
On 4× L40S 48GB at AWQ 4-bit. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K471.31 GB✓ fits
8K232.62 GB✓ fits
33K510.47 GB✓ fits
66K220.94 GB✓ fits
131K141.88 GB✓ fits
Other GPUs that run Mistral Large 3 (256K)
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB8×17135$0.0648
A100 80GB SXM42×1195$0.5578
H100 SXM5 80GB2×2002$0.6937
AMD MI300X 192GB1×931.6$0.7454
H200 SXM5 141GB2×2869$0.7747
AMD MI325X 288GB1×1055$0.9218
Frequently asked questions

Can Mistral Large 3 (256K) run on L40S 48GB?

Yes, but it needs 4× L40S 48GB. At AWQ 4-bit the total footprint is 150.2 GB (129.3 GB of weights plus 20.9 GB of KV cache at 8K × 8 concurrent users), against 192 GB of available VRAM across 4 GPUs.

How fast is Mistral Large 3 (256K) on L40S 48GB?

43.03 tokens/sec per stream and 2065 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 6.40 s, and inter-token latency is 23.2 ms.

What does Mistral Large 3 (256K) on L40S 48GB cost per million tokens?

$0.3766 per 1M output tokens at 4× L40S 48GB ($0.7000/hr). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does Mistral Large 3 (256K) need on L40S 48GB?

129.3 GB for the weights at AWQ 4-bit, plus 20.9 GB of KV cache for 8 concurrent requests at 8K context — 150.2 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is L40S 48GB the right choice for Mistral Large 3 (256K)?

It works. Compare $0.3766 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 8× RTX 5090 32GB at $0.0648, and the fastest is 4× Cerebras CS-3 (WSE-3) at 47811 tok/s.

Adjust the workload yourself

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

Open in calculator