Google Gemma 3 12B vs Microsoft Phi 4
Side by side from live data: prices, limits, who serves each model and how reliably, and how they do on our chess arena.
In short
- Gemma 3 12B is 1.2× cheaper at a typical mix of 3 input tokens per output token.
- Gemma 3 12B accepts longer prompts: 131K vs 16K tokens.
- Gemma 3 12B is the newer model (released 2025-03-13).
| Gemma 3 12B | Phi 4 | |
|---|---|---|
| Vendor | Microsoft | |
| Released | 2025-03-13 | 2025-01-10 |
| Input / 1M | $0.05 | $0.07 |
| Output / 1M | $0.15 | $0.14 |
| Cached input / 1M | — | — |
| Blended / 1M (3:1) | $0.075 | $0.088 |
| Context | 131K | 16K |
| Max output | 16K | 15K |
| Accepts | text, image | text |
| Reasoning | No | No |
| Providers on OpenRouter | 4 | 2 |
| Cheapest provider, output | $0.15 | $0.14 |
| Best provider uptime, 24 h | 100% | 100% |
| Chess arena rating | not playing | not playing |
| Vendor API status | Operational | — |
Cost calculator
Gemma 3 12B
$5.25 per month$0.17 per day
Phi 4
$6.30 per month$0.21 per day
Gemma 3 12B saves $1.05 a month on this workload.
At list prices per 1M tokens; prompt caching and batch discounts are not included.
Compare Gemma 3 12B with