Includes 694 hourly rounds up to 2026-11-16 from NIMStats by Mauro Druwel (source), measured with a different prompt and timeout. Our own probe adds a round every 30 minutes.
Model comparison
Head-to-head from live probes, last 24 hours.
Head-to-head stats
muse-glimmer-30b leads 3 rows, step-5-preview leads 4.
| muse-glimmer-30b | Metric | step-5-preview |
|---|---|---|
| meta Serving | Status | Unstable stepfun |
| 64● | Score | 45 |
| 91.7%● | Uptime | 50.0% |
| 33.4 s | First token (median) | ●7.49 s |
| 61.1 s | Full reply (median) | ●25.4 s |
| 26.5 s | Best reply | ●7.40 s |
| 8.2 tok/s | Throughput (median) | ●10.7 tok/s |
| 0 | Round wins | 0 |
| #9● | Leaderboard rank | #10 |
| 33.3% | Faster when both answered (6 rounds) | 66.7% |
marks the better value in each row
Full reply time overlay
Each probe, both models on one axis. Lower is better.
Win timeline
Per round, oldest left: which model replied faster.
BAAA·BAAABBA
Amuse-glimmer-30b Bstep-5-preview ·neither answered