Model

Turing-NLG (17B)

CLOSED

Microsoft · Turing-NLG family · released Feb 13, 2020

At launch the largest Transformer LM (17B); showcased DeepSpeed/ZeRO; research-only, weights never released.

ReasoningCodingVisionFunction callingTool useAgentic

-449.6

Elo · unrated

Parameters

17B

Active params

17B (dense)

Context

1K tokens

Architecture

Dense Transformer decoder (78 layers, hidden size 4256, 28 attention heads)

License

Proprietary (weights never released)

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

LAMBADAReasoning68.0%#22

best: GPT-3 175B · 86.4%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

—

Output / M tok

—

Turing-NLG family

Elo progression across releases

Resources

API price — · each benchmark row carries its own source badge (see methodology)