Gemma 4 31B (Non-reasoning) vs Trinity Large Thinking
Which is better for the money: Gemma 4 31B (Non-reasoning) by Google or Trinity Large Thinking by Arcee AI? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-08-19.
Clear winner
Verdict
Pick Gemma 4 31B (Non-reasoning)
Trinity Large Thinking vs Gemma 4 31B (Non-reasoning): 16.1% less intelligence, 652.7% more cost to run the AA benchmark, and 88.5% slower end-to-end.
| Metric | Gemma 4 31B (Non-reasoning) | Trinity Large Thinking | Difference |
|---|---|---|---|
| Intelligence (higher is better) | 22.3 | 18.7 | −16.1% Trinity Large Thinking is 16.1% less intelligent |
| Total AA cost (lower is better) | $33.84 | $254.72 | +652.7% Trinity Large Thinking is 7.5× more expensive |
| End-to-end time (lower is better) | 7.47s | 14.08s | +88.5% Trinity Large Thinking is 88.5% slower |
| Price per 1M tokens (in / out) | $0.15 / $0.40 | $0.23 / $0.88 | |
| Input modalities | Not reported | Not reported |
Today's Pareto frontier
Against the frontier
- Gemma 4 31B (Non-reasoning) is off the frontier: at the same or lower cost, GPT-5.6 Luna (medium) scores 38.90 for $21.1.
- Trinity Large Thinking is off the frontier: at the same or lower cost, GPT-5.6 Luna (max) scores 52.30 for $172.
Gemma 4 31B (Non-reasoning)Trinity Large ThinkingToday's Pareto frontierOpen-weights frontierEvery other model
Matchup
Compare something else
Agents:
Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/gemma-4-31b-non-reasoning/trinity-large-thinking · Markdown: /api/v1/compare/gemma-4-31b-non-reasoning/trinity-large-thinking?format=md · llms.txt · agent skill