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
$0.2500
input / 1M
▼ −44% since first
$1.50
output / 1M
▼ −44% since first
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=4
Last 30 days
100.0%
n=312
Median response time
1,326ms
n=312
Based on 332 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)
312
OK responses (30d)
312
Total calls (7d)
4
OK responses (7d)
4
Tokonomix benchmark verdicts
Quality decline across categories with reasoning performance now unmeasured
Gemini 3.1 Flash Lite shows a notable 6-point drop in overall quality score, falling from 99.3 to 93.3 out of 100 in the current benchmark window. The model maintains perfect scores in coding and multilingual tasks at 100 each, but creative performance registered at just 80, suggesting potential regression in generative capabilities. Most concerning is the complete absence of reasoning scores in the current window, despite achieving a perfect 100 in this category previously. This missing data point makes it difficult to assess whether the model has actually lost reasoning capability or if the test coverage has simply changed. Latency remains relatively stable with a marginal increase from 1408ms to 1460ms at the median, representing a 52ms degradation that should be negligible for most use cases. The consistency in test runs at 5 per window provides reasonable confidence in these measurements. Users should be aware that while specialized tasks like coding and multilingual processing remain strong, the overall reliability appears to have decreased. The missing reasoning benchmark is particularly notable given its previous perfect performance, and users relying on logical inference capabilities should exercise caution until this metric is re-established.
Quality
93.3
Latency p50
1,460 ms
Test runs
5