Nemotron-3-Nano-Omni-Q4_K_M-NVIDIA Hermes Benchmark Failure Report

Result: Failed by user request due to speed

This run was stopped because the model was too slow for the benchmark. It never reached the first required tool call. Hermes waited 467.0 seconds on the initial model response before I interrupted it, then a direct continuation attempt also made no tool call within 29.9 seconds.

Model label
Nemotron-3-Nano-Omni-Q4_K_M-NVIDIA
LM Studio identifier
nvidia/nemotron-3-nano-omni
Hermes displayed model
nemotron-3-nano-omni
Quantization noted by user
Q4_K_M
Reasoning effort
high
LM Studio context
262,144 tokens
Session key
20260527_061702_0962bc
Hermes session duration
8m 54s

Run Summary

Checkpoint Status Evidence
Confirm loaded model Passed LM Studio showed nvidia/nemotron-3-nano-omni loaded; Hermes displayed nemotron-3-nano-omni.
Set reasoning high Passed Hermes confirmed Reasoning effort set to 'high'.
Initial prompt processing Slow LM Studio spent several minutes in PROCESSINGPROMPT before switching to GENERATING.
Read test-prompt.md Failed No read_file call was made. Hermes exited with 0 tool calls.
Create benchmark HTML Failed No benchmark artifact was written for this model.

Timeline

Time Event
2026-05-27 06:17:02 +0600 Fresh Hermes session opened. Session key: 20260527_061702_0962bc.
Start of run LM Studio confirmed nvidia/nemotron-3-nano-omni, size 26.10 GB, context 262144, parallel 4, device MGPC.
Initial model call LM Studio moved from PROCESSINGPROMPT to GENERATING, but Hermes received no tool call or usable response.
After 467.0 seconds I interrupted the first model call because it had not even reached the required prompt-file read.
Continuation attempt I instructed Hermes to immediately call the file-reading tool and write the required HTML.
After 29.9 more seconds User requested failure because the model was too slow; I interrupted and exited Hermes.
2026-05-27 06:26:02 +0600 Failure report written. LM Studio had returned to IDLE.

Tool Call Audit

Tool Count Detail
Hermes tools 0 Hermes session summary: Messages: 1 (1 user, 0 tool calls).
Expected first tool 0 read_file for /Users/armanshawon/Documents/Benchmark/test-prompt.md was never reached.
File writes 0 No model-generated benchmark file or partial report was created.

Failure Judgment

This model should be recorded as failed for this Hermes benchmark on this machine. The failure reason is not a template error or tool crash; it is insufficient practical speed and no observable task progress. Even with LM Studio reporting active generation, Hermes reached neither the prompt-file read nor any output file write before the run was cancelled.

Raw Notes

Hermes session: 20260527_061702_0962bc
Hermes displayed model: nemotron-3-nano-omni
LM Studio identifier: nvidia/nemotron-3-nano-omni
Reasoning: high
Context: 262144
Initial wait before first interruption: 467.0s
Second wait before user-requested failure: 29.9s
Hermes summary: Duration 8m 54s; Messages 1; Tool calls 0
Final LM Studio status: IDLE