Me
llama-3.3-70b-instruct-fp8-fast Benchmark & Insights
Meta Cloudflare Workers AI
Updated Jun 6, 2026 All models
Sample size
113 runs
in window
Accuracy
83.5%
consensus match · 115d
Confidence
88%
over 115 runs
Window end
Jun 6, 2026
most recent run
Input price
$0.29/MTok
prompt tokens
Output price
$2.25/MTok
completion tokens
Model insights
- 01 Erratic rather than biased: 13 overratings against 10 underratings, swinging between "low" and "high" with 7 false "unsafe" calls across the longest record (113 days).
- 02 Inconsistency in both directions makes it less predictable than models with a clear tilt.