GPT-5
OpenAI · moe · 500B parameters · 400,000 context
Parameters
500B
Context Window
391K tokens
Architecture
MoE
Best GPU
B200 NVL (pair)
Cheapest API
$10.00/M
Intelligence Brief
GPT-5 is a 500B parameter Mixture-of-Experts (16 experts, 2 active) model from OpenAI, featuring Grouped Query Attention (GQA) with 120 layers and 12,288 hidden dimensions. With a 400,000 token context window, it supports tools, vision, structured output, code, math, multilingual, reasoning. The most cost-effective API deployment is via openai at $10.00/M output tokens. For self-hosted inference, B200 NVL (pair) delivers optimal throughput at $39858/month.
Provider pricing
2 providers · canonical: openai| Provider | Input $/M | Output $/M ▲ | Notes |
|---|---|---|---|
| openaicanonical | $1.25 | $10.00 | cheapest input · cheapest output |
| openrouter | $1.25 | $10.00 | cheapest input · cheapest output |
Prices update via the nightly pricing cron + admin approvals at /admin/ingest-queue. The leaderboard's Input/Output cells show the canonical rate above; this table shows the full spread.
Recent changes
Loading…
Related models
5 suggestions- GPT-4.5 PreviewGPT · 300B$150.00/M out
- GPT-5.5GPT · 110B$30.00/M out
- GPT-5 MiniGPT · 25B$2.00/M out
- GPT-5 NanoGPT · 4B$0.400/M out
- GPT-4 TurboGPT-4 · 50B$30.00/M out
Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.
Architecture Details
Memory Requirements
BF16 Weights
1000.0 GB
FP8 Weights
500.0 GB
INT4 Weights
250.0 GB
Fits on (single GPU) — most practical first
Fits on (multi-GPU with Tensor Parallelism)
Multi-GPU configurations use Tensor Parallelism (TP) to split model layers across GPUs. Requires NVLink or NVSwitch interconnect for optimal performance.
GPU Compatibility Matrix
GPT-5 is compatible with 2% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
BF16 · 4 GPUs · tensorrt-llm
88/100
score
Throughput
140.0 tok/s
Latency (ITL)
7.1ms
Est. TTFT
1ms
Cost/Month
$39858
Cost/M Tokens
$108.33
BF16 · 8 GPUs · tensorrt-llm
83/100
score
Throughput
140.0 tok/s
Latency (ITL)
7.1ms
Est. TTFT
1ms
Cost/Month
$34088
Cost/M Tokens
$92.65
BF16 · 8 GPUs · tensorrt-llm
83/100
score
Throughput
140.0 tok/s
Latency (ITL)
7.1ms
Est. TTFT
1ms
Cost/Month
$34164
Cost/M Tokens
$92.86
Deployment Options
API Deployment
openai
$10.00/M
output tokens
Single GPU
Requires multi-GPU setup (500 GB VRAM needed)
Multi-GPU
B200 NVL (pair) x4
140.0 tok/s
TP· $39858/mo
API Pricing Comparison
| Provider | Input $/M | Output $/M | Badges |
|---|---|---|---|
| openai | $1.25 | $10.00 | Cheapest |
| openrouter | $1.25 | $10.00 |
Cost Analysis
| Provider | Input $/M | Output $/M | ~Monthly Cost |
|---|---|---|---|
| openaiBest Value | $1.25 | $10.00 | $56 |
| openrouter | $1.25 | $10.00 | $56 |
Cost per 1,000 Requests
Short (500 tok)
$2.63
via openai
Medium (2K tok)
$10.50
via openai
Long (8K tok)
$30.00
via openai
Performance Estimates
Throughput by GPU
VRAM Breakdown (B200 NVL (pair), BF16)
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy GPT-5
Similar Models
GPT-5.5
700B params · moe
Quality: 50
from $30.00/M
GPT-4.5 Preview
1500B params · moe
Quality: 93
from $150.00/M
Claude Opus 4.7
500B params · moe
Quality: 90
from $25.00/M
Grok 4.3
500B params · moe
Quality: 50
from $2.50/M
Snowflake Arctic 480B
480B params · moe
Quality: 50
from $1.50/M
Frequently Asked Questions
How much VRAM does GPT-5 need for inference?
GPT-5 requires approximately 1000.0 GB of VRAM at BF16 precision, 500.0 GB at FP8, or 250.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (204800 bytes per token) and activations (~4.00 GB).
What is the best GPU for GPT-5?
The top recommended GPU for GPT-5 is the B200 NVL (pair) (x4) using BF16 precision. It achieves approximately 140.0 tokens/sec at an estimated cost of $39858/month ($108.33/M tokens). Score: 88/100.
How much does GPT-5 inference cost?
GPT-5 API inference starts from $1.25/M input tokens and $10.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.