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
How this model compares to the rest of the field on each prompt category, from a pairwise fit over the same prompts. The raw judge score sits underneath each number.
Win rate per category: how often this model beats a field-average model on a prompt from that category. 50% is average, not a failing grade. It is not a percentage of correct answers.
Pricing
What you pay per million tokens when you use this model on Tokonomix, plus an estimate for a typical conversation.
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
100.0%
n=6
Last 30 days
100.0%
n=6
Median response time
50,588ms
n=6
Based on 391 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)
6
OK responses (30d)
6
Total calls (7d)
6
OK responses (7d)
6
Tokonomix benchmark verdicts
Qwen3-32B drops to 56.8 quality, latency improves 17% to 18.0s
Qwen3-32B shows a mixed performance shift in this benchmark window, with overall quality declining slightly from 58.0 to 56.8 while latency improved meaningfully from 21.7s to 18.0s. The 17% latency improvement brings response times closer to acceptable levels, though they remain notably high for a 32B parameter model. Category performance reveals dramatic swings: reasoning capability jumped impressively to 89 from a previous coding score of 52, and coding maintained strong performance at 72. However, factual accuracy collapsed to just 10, a concerning regression that suggests significant reliability issues with knowledge-based queries. The previous window's multilingual strength at 91 and creative capabilities are not represented in current testing categories, making direct comparison incomplete. This benchmarking period captures only 5 test runs, matching the previous window's sample size. Users should exercise caution with factual queries given the severe accuracy drop, while those prioritizing reasoning tasks may find improved utility. The latency gains are welcome but the model still requires substantial patience per request. Overall trajectory remains uncertain given the quality decline despite operational improvements.
Quality
56.8
Latency p50
17,953 ms
Test runs
5