GLM 5.1
z-ai/glm-5.1
44.3
median tok/s
Throughput Runs
9
TTFT Runs
6
Avg TTFT
2394ms
Avg Throughput
45.9 tok/s
Total Cost
$0.0750
Commentary
by derivedGLM 5.1 posts 44.3 tok/s median throughput with 2353ms median TTFT across 15 runs. Success rate is 100.0% with $0.08 total spend.
Startup latency on ttft-factual landed in the 2353ms range.
Startup latency on ttft-definition landed in the 2353ms range.
Sustained decode speed on throughput-data-structures contributed to the 44.3 tok/s median.
Sustained decode speed on throughput-api-design contributed to the 44.3 tok/s median.
Sustained decode speed on throughput-essay contributed to the 44.3 tok/s median.
Notable Prompts
Fastest throughput run peaked at 66.0 tok/s.
Slowest startup path took 3170ms to first token.
All Runs
| Prompt | Type | Tok/s | TTFT | Tokens | Cost | |
|---|---|---|---|---|---|---|
1. Api Design throughput-api-design | throughput | 60.2 | 6579ms | 2590 | $0.0084 | |
2. Api Design throughput-api-design | throughput | 47.9 | 4573ms | 2616 | $0.0085 | |
3. Api Design throughput-api-design | throughput | 44.3 | 4454ms | 3042 | $0.0098 | |
1. Data Structures throughput-data-structures | throughput | 66.0 | 2245ms | 2400 | $0.0078 | |
2. Data Structures throughput-data-structures | throughput | 52.8 | 4269ms | 2732 | $0.0089 | |
3. Data Structures throughput-data-structures | throughput | 40.8 | 3927ms | 2606 | $0.0085 | |
1. Essay throughput-essay | throughput | 37.3 | 5535ms | 2190 | $0.0071 | |
2. Essay throughput-essay | throughput | 30.4 | 3911ms | 2586 | $0.0084 | |
3. Essay throughput-essay | throughput | 33.5 | 3962ms | 2184 | $0.0071 | |
1. Definition ttft-definition | ttft | n/a | 3170ms | 53 | $0.0002 | |
2. Definition ttft-definition | ttft | n/a | 2530ms | 46 | $0.0002 | |
3. Definition ttft-definition | ttft | n/a | 3071ms | 53 | $0.0002 | |
1. Factual ttft-factual | ttft | n/a | 2175ms | 12 | $0.0001 | |
2. Factual ttft-factual | ttft | n/a | 2150ms | 12 | $0.0001 | |
3. Factual ttft-factual | ttft | n/a | 1268ms | 12 | $0.0001 |
15 runs · Throughput rows require valid long-output runs · TTFT shown for all successful runs