LLMHOSTING.AI the compute price terminal

Llama 3.1 405B API pricing

6 providers serve Llama 3.1 405B. Llama 3.1 405B is an open-weight model (Llama 3.1 Community) with 405B total parameters, up to 128K context. Prices below are per 1M tokens, cheapest input first (self-hosting math further down).

Refreshed 2026-08-15 · list prices from provider APIs
Input floor $0.120/1M at Hyperbolic Output floor $0.300/1M at Hyperbolic Find your workload winner →
Provider Input $/1M Output $/1M Cache read $/1M Context
Hyperbolic $0.120 $0.300output floor - 33K Use
Nebius $1.00 $3.00 - 128K Use
SambaNova $5.00 $10.00 - 16K Use
Google Vertex AI $5.00 $16.00 - 128K Use
Azure AI Foundry $5.33 $16.00 - 128K Use
Oracle OCI $10.68 $10.68 - 128K Use

The cheapest way to run Llama 3.1 405B

On an illustrative 70M-input / 30M-output monthly workload, today's lowest listed cost is roughly $17.40/month, at Hyperbolic. Input-heavy, output-heavy and cache-heavy workloads can produce different winners. Use the live mix calculator below rather than combining floors from two different providers. On output-token cost alone, the API floor remains below our self-hosting estimate of ~$6.17/1M. Details below. Run your own numbers in the breakeven calculator, or read the breakeven math.

Your actual monthly cost

Uses today's list prices

Cache read uses the listed cache rate where available; otherwise it falls back to normal input price. Batch, write-cache, volume and negotiated discounts are excluded.

Or run it yourself: Llama 3.1 405B self-hosting economics

Llama 3.1 405B is open-weight (Llama 3.1 Community), so the API price above competes with the GPU-hour market. 405B total parameters needs roughly 557 GB VRAM at FP8 or 279 GB at INT4, KV-cache headroom included.

No single GPU fits Llama 3.1 405B at FP8 (557 GB needed), so it needs a multi-GPU node. Cheapest tracked option: 3× MI300X (192 GB each) at roughly $1.50/hr total (RunPod). Rent 3× MI300X at RunPod → INT4 quantization drops the requirement to 279 GB. Tune it in the calculator.
Breakeven estimate: a well-batched vLLM deployment on 3× MI300X ($1.50/hr) at ~135 aggregate tok/s and 50% utilization works out to roughly $6.17 per 1M output tokens, versus $0.300 via the cheapest output-token API (Hyperbolic). The API is cheaper unless your utilization or throughput beats these assumptions. Planning estimate: throughput varies with hardware, quantization, batch size and context. Tune it in the calculator.

Related models

ModelCheapest in $/1MCheapest out $/1MProviders
Llama 3.3 70B $0.100 $0.200 15
Llama 3.1 8B $0.020 $0.030 15
Llama 4 Scout $0.050 $0.100 11
Llama 3.1 70B Instruct $0.120 $0.300 10
Llama 4 Maverick $0.050 $0.100 9
Llama 3.2 3B $0.020 $0.020 9

FAQ

What is the cheapest API for Llama 3.1 405B?

As of 2026-08-15, the lowest input price for Llama 3.1 405B is Hyperbolic at $0.120 per 1M input tokens. The lowest output price is Hyperbolic at $0.300 per 1M output tokens. The cheapest provider for a real workload depends on its input/output mix. The next-lowest input price is Nebius at $1.00, a 733% difference.

How much VRAM do you need to self-host Llama 3.1 405B?

Llama 3.1 405B has 405B parameters, so plan for roughly 557 GB of VRAM at FP8 or 279 GB at INT4/AWQ, KV-cache headroom included. That exceeds a single GPU. A typical node is 3× MI300X.

← All models · GPU rental prices · Breakeven calculator · Get this data as JSON