SciGLM 6B
Tsinghua · dense · 6.2B parameters · 8,192 context
Parameters
6.2B
Context Window
8K tokens
Architecture
Dense
Best GPU
A10G
Intelligence Brief
SciGLM 6B is a 6.2B parameter DENSE model from Tsinghua, featuring Grouped Query Attention (GQA) with 28 layers and 4,096 hidden dimensions. With a 8,192 token context window, it supports math, reasoning. For self-hosted inference, A10G delivers optimal throughput at $285/month.
Architecture Details
Memory Requirements
BF16 Weights
12.4 GB
FP8 Weights
6.2 GB
INT4 Weights
3.1 GB
GPU Compatibility Matrix
SciGLM 6B is compatible with 95% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
BF16 · 1 GPU · vllm
100/100
score
Throughput
261.3 tok/s
Latency (ITL)
3.8ms
Est. TTFT
1ms
Cost/Month
$285
Cost/M Tokens
$0.41
BF16 · 1 GPU · vllm
100/100
score
Throughput
406.3 tok/s
Latency (ITL)
2.5ms
Est. TTFT
0ms
Cost/Month
$332
Cost/M Tokens
$0.31
BF16 · 1 GPU · vllm
100/100
score
Throughput
438.9 tok/s
Latency (ITL)
2.3ms
Est. TTFT
0ms
Cost/Month
$370
Cost/M Tokens
$0.32
Deployment Options
API Deployment
No API pricing available
Single GPU
A10G
$285/mo
Min VRAM: 6 GB
Multi-GPU
RTX 3080 x2
503.8 tok/s
TP· $266/mo
API Pricing Comparison
No API pricing data available for this model.
Performance Estimates
Throughput by GPU
VRAM Breakdown (A10G, BF16)
Precision Impact
bf16
12.4 GB
weights/GPU
~261.3 tok/s
fp8
6.2 GB
weights/GPU
int4
3.1 GB
weights/GPU
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy SciGLM 6B
Self-Hosted Infrastructure
Similar Models
ChatGLM3 6B
6B params · dense
Quality: 50
Yi 6B 200K
6B params · dense
Quality: 50
DeepSeek Coder 6.7B
6.7B params · dense
Quality: 50
from $0.20/M
MPT 7B
6.7B params · dense
Quality: 36
StarCoder2 7B
6.73B params · dense
Quality: 35
from $0.15/M
Frequently Asked Questions
How much VRAM does SciGLM 6B need for inference?
SciGLM 6B requires approximately 12.4 GB of VRAM at BF16 precision, 6.2 GB at FP8, or 3.1 GB at INT4 quantization. Additional VRAM is needed for KV-cache (14336 bytes per token) and activations (~0.60 GB).
What is the best GPU for SciGLM 6B?
The top recommended GPU for SciGLM 6B is the A10G using BF16 precision. It achieves approximately 261.3 tokens/sec at an estimated cost of $285/month ($0.41/M tokens). Score: 100/100.
How much does SciGLM 6B inference cost?
SciGLM 6B inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.