Gemma
173 opinions tracked this week · 7-day rolling window
30-day experience score
Version breakdown
Only comments that explicitly name a version contribute to its score.
| Gemma 4 31B | 61.4−4.7 | n=18 | Confidence: Medium | |
|---|---|---|---|---|
| Gemma 4 26B A4B | 53.1−13.0 | n=13 | Confidence: Medium | |
| Gemma 2 27B | mentions only | n=1 | Confidence: Insufficient | |
| Gemma 3 12B | mentions only | n=1 | Confidence: Insufficient | |
| Gemma 4 31B (free) | mentions only | n=1 | Confidence: Insufficient |
139 generic family mentions are included in the family score, not any version score.
By experience dimension
| General text | 64 | n=56 | |
|---|---|---|---|
| Coding | 68 | n=19 | |
| Reasoning | 73 | n=28 | |
| Image / vision | not enough data | n=0 | |
| Roleplay / creative | 71 | n=6 | |
| Safety & refusals | not enough data | n=1 | |
| Speed & latency | 52 | n=47 | |
| Local deploy | 61 | n=16 |
Top positive drivers
- strong reasoning Reasoning ×19
- speed quality Speed & latency ×18
- speed quality Coding ×4
- competitor flight General text ×4
- creative surprise General text ×4
Top negative drivers
- performance issue Speed & latency ×10
- speed quality Speed & latency ×6
- performance issue General text ×5
- competitor flight General text ×4
- tool failure General text ×3
What users are saying
🟢 “I can send gemma e4b narrated audio and it will do an amazing job of validating it, and making corrective suggestions on phoenetics on an e4b model back to my renderer, then loop 3 times to correct and I'll be damned if it almost always fix”
via reddit · text_general · en
🟢 “Gemma 4 E2B are both astounding for their tiny size”
via reddit · reasoning · en
🟢 “12B actually surpassed my expectations. I assumed it wouldn't work at all, but it... kinda does.”
via reddit · coding · en
🟢 “Gemma 12B is such good value for money for agentic RPGs and general writing.”
via reddit · speed_latency · en
⚪ “I'm very happy with Gemma 4 12B, the QAT model with MTP enabled rips on my 3080 Ti. The only problem I have is tool calling, it will frequently stall on tool calls for more complex tasks. Other than that it's great for a day to day assistan”
via reddit · text_general · en