Model Detail
Gemma 4 31B
The better local model when quality matters more than speed and concurrency.
Benchmark score
92/100
Average latency
22.59s
Role
Escalation local model
Strengths
- Best quality in the Gemma local pair
- Better routing judgment on the benchmark pack
- Cleaner concise answers overall
Weaknesses
- Much slower average latency
- Worse fit for high-concurrency local loops
- Still wrapped strict JSON in code fences
Operator read
Stronger output quality and better routing judgment than 26B, but roughly 3x slower in this quick benchmark pack.
Source artifacts
Raw machine-readable files for anyone who wants to dig deeper or run their own analysis.