Published reference2026-09-09

IFBench

Source-reported protocol · sample size not reported. Compare only matched task sets and protocols.

GLM-5.3-Flash NVFP4 · Published reference · NVIDIA · Reported evaluation

↗ Published result

NVIDIA

Protocol as described by the publisher

nvidia/GLM-5.3-Flash-NVFP4 model card ↗

Reported score60.5%

Throughput—

Tokens per second · see measurement basis

Concurrency—

As recorded on this receipt

Published reference. This is a source-reported result, not a test run by this lab. NVIDIA published NVFP4 0.6054 (60.54). Same table lists BF16 IFBench 0.613 as a baseline only. temperature=1.0, top_p=0.95, max_new_tokens=327680. Recipe: ~3.33× smaller vs BF16 (Model Optimizer NVFP4).

Run configuration & provenance +

Context: Not recorded
Temperature: Not recorded
Wall time: Not recorded
Weight file: Provider-managed / not reported
Revision: Not reported
Reasoning: Not reported
Agent version: Not reported
Output token limit: Not reported
Dataset revision: Not reported
Harness host: Not reported
Model host: Published reference
Input tokens: Not reported
Output tokens: Not reported
Cached input tokens: Not reported
Agent turn / step limit: Not reported
Concurrency: Not reported
Weight file size: Not reported

Throughput measurement: Measurement method not separately reported. This should not be treated as standardized decode-only speed.

Provider evaluation

NVIDIA published NVFP4 0.6054 (60.54). Same table lists BF16 IFBench 0.613 as a baseline only. temperature=1.0, top_p=0.95, max_new_tokens=327680. Recipe: ~3.33× smaller vs BF16 (Model Optimizer NVFP4).

Run ID: research-glm-5.3-flash-nvfp4-published-reference-ifbench

Tasks

Task explorer

Passed tasks first, then other outcomes grouped by recorded issue. Use the filters to inspect a particular outcome.

0 task records

↗

No task records in this receipt.

The headline or summary is available above. Individual outcomes have not been provided.

Explore test coverage →