VILA 1.5 40B
NVIDIA · dense · 40B parameters · 8,192 context
Parameters
40B
Context Window
8K tokens
Architecture
Dense
Best GPU
H100 SXM
Cheapest API
$1.00/M
Quality Score
73/100
Intelligence Brief
VILA 1.5 40B is a 40B parameter DENSE model from NVIDIA, featuring Grouped Query Attention (GQA) with 48 layers and 8,192 hidden dimensions. With a 8,192 token context window, it supports vision, structured output, code, math, multilingual. On standardized benchmarks, it achieves MMLU 74, HumanEval 42, GSM8K 78. The most cost-effective API deployment is via nvidia-nim at $1.00/M output tokens. For self-hosted inference, H100 SXM delivers optimal throughput at $1794/month.
Provider pricing
1 provider · canonical: nvidia-nim| Provider | Input $/M | Output $/M ▲ | Notes |
|---|---|---|---|
| nvidia-nimcanonical | $1.00 | $1.00 | 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- VILA 1.5 13BVILA · 13B$0.300/M out
- VILA 1.5 3BVILA · 3B$0.080/M out
HelpSteer2 Llama 3.1 70BLlama 3.1 · 70.6B$0.400/M out- Llama 3.1 Nemotron 51BLlama 3.1 · 51B$0.400/M out
- Llama 3.1 Nemotron 70B InstructLlama 3.1 · 70.6B$0.880/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
80.0 GB
FP8 Weights
40.0 GB
INT4 Weights
20.0 GB
GPU Compatibility Matrix
VILA 1.5 40B is compatible with 52% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
FP8 · 1 GPU · tensorrt-llm
100/100
score
Throughput
700.6 tok/s
Latency (ITL)
1.4ms
Est. TTFT
0ms
Cost/Month
$1794
Cost/M Tokens
$0.97
FP8 · 1 GPU · tensorrt-llm
100/100
score
Throughput
418.3 tok/s
Latency (ITL)
2.4ms
Est. TTFT
0ms
Cost/Month
$1794
Cost/M Tokens
$1.63
FP8 · 1 GPU · tensorrt-llm
100/100
score
Throughput
836.6 tok/s
Latency (ITL)
1.2ms
Est. TTFT
0ms
Cost/Month
$940
Cost/M Tokens
$0.43
Deployment Options
API Deployment
nvidia-nim
$1.00/M
output tokens
Single GPU
H100 SXM
$1794/mo
Min VRAM: 40 GB
Multi-GPU
A100 80GB SXM x2
271.7 tok/s
TP· $2259/mo
API Pricing Comparison
| Provider | Input $/M | Output $/M | Badges |
|---|---|---|---|
| nvidia-nim | $1.00 | $1.00 | Cheapest |
Cost Analysis
| Provider | Input $/M | Output $/M | ~Monthly Cost |
|---|---|---|---|
| nvidia-nimBest Value | $1.00 | $1.00 | $10 |
Cost per 1,000 Requests
Short (500 tok)
$0.70
via nvidia-nim
Medium (2K tok)
$2.80
via nvidia-nim
Long (8K tok)
$10.00
via nvidia-nim
Performance Estimates
Throughput by GPU
VRAM Breakdown (H100 SXM, FP8)
Precision Impact
bf16
80.0 GB
weights/GPU
fp8
40.0 GB
weights/GPU
~700.6 tok/s
int4
20.0 GB
weights/GPU
Quality Benchmarks
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy VILA 1.5 40B
Self-Hosted Infrastructure
Similar Models
VILA 1.5 13B
13B params · dense
Quality: 62
from $0.30/M
Falcon 40B
40B params · dense
Quality: 48
from $0.80/M
Phi 3.5 MoE
41.9B params · moe
Quality: 74
from $0.30/M
Aya 23 35B
35B params · dense
Quality: 50
from $1.50/M
Command R (August 2024)
35B params · dense
Quality: 68
from $0.60/M
Frequently Asked Questions
How much VRAM does VILA 1.5 40B need for inference?
VILA 1.5 40B requires approximately 80.0 GB of VRAM at BF16 precision, 40.0 GB at FP8, or 20.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (196608 bytes per token) and activations (~3.00 GB).
What is the best GPU for VILA 1.5 40B?
The top recommended GPU for VILA 1.5 40B is the H100 SXM using FP8 precision. It achieves approximately 700.6 tokens/sec at an estimated cost of $1794/month ($0.97/M tokens). Score: 100/100.
How much does VILA 1.5 40B inference cost?
VILA 1.5 40B API inference starts from $1.00/M input tokens and $1.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.