Speed analysis
Latency measured across all benchmark runs. P50 (median) and P95 (95th percentile) give a realistic picture of response speed under normal and peak load.
Quality scores
Evaluation results from judge-model scoring across diverse task categories. Scores reflect coherence, accuracy and instruction-following.
Pricing history
Direct provider rates per million tokens, plus a typical-conversation cost estimate.
Pricing over time
Input & output per 1M tokens · step-line = price changes
$0.9100
input / 1M
— stable
$0.9100
output / 1M
— stable
Tokens per second
Throughput in tokens per second, derived from measured P50 latency. Higher is better; fluctuations track provider-side load.
Estimated from P50 latency × 200 output tokens — the absolute number depends on this assumption; the trend is what matters.
Capabilities
Availability
Availability
How often this model answers when we call it — measured across real API requests and live tests over the last 30 days. This is separate from quality: these numbers only tell you whether the model responds, not how good the answer is.
Last 7 days
—
Last 30 days
100.0%
n=36
Median response time
4,186ms
n=36
Based on 426 measurements over the last 30 days.
Technical details
Only live API calls and live-test requests count — internal probes and benchmark runs are excluded.
Calls with a custom API key (BYOK) are excluded: those failures are key-specific, not a sign of model downtime.
Failed calls are NOT included in quality scores — quality is measured on successful responses only. Availability and quality are independent signals.
Median response time (p50) across successful calls with a recorded duration. Outliers (very slow or very fast calls) pull the median less than the average.
Total calls (30d)
36
OK responses (30d)
36
Total calls (7d)
0
OK responses (7d)
0
Tokonomix benchmark verdicts
Quality drops 10 points with slower response times and factual weakness
Qwen2.5-VL-72B-Instruct experienced a notable decline in this benchmark window, with overall quality falling from 98.8 to 88.8 points while latency increased by 37 percent to 8.9 seconds at the median. The model demonstrates exceptional strength in multilingual and reasoning tasks, both scoring perfect 100s, and maintains outstanding creative capabilities at 98 points. However, factual accuracy emerged as a significant weakness, scoring just 57 points and representing a substantial gap in the model's performance profile. The previous window included coding benchmarks where the model scored 98, but this category was not tested in the current window, making direct comparison incomplete. Thelatency increase from 6.5 to 8.9 seconds suggests either infrastructure changes or increased processing complexity. Despite the quality decline, the model retains strong capabilities in three of four tested categories. Users should be aware of the factual accuracy limitations when deploying this model for knowledge-intensive applications, while the model remains well-suited for creative, multilingual, and reasoning-heavy workloads where it continues to excel.
Quality
88.8
Latency p50
8,876 ms
Test runs
5