Compare / Gemma 3 4B vs Ministral 3 8B 2512
Gemma 3 4B vs Ministral 3 8B 2512
Which is cheaper, which is more capable, and which does better at each kind of work, from 15 results on 3 sources that measured both models.
Gemma 3 4B or Ministral 3 8B 2512? The short answer
- Gemma 3 4B is cheaper: $0.062 against $0.15 per million tokens, 2.4 times less.
- They score about the same on the Artificial Analysis Intelligence Index (4.8 and 5.5).
- On our business benchmarks, Gemma 3 4B ranks higher for sticking to the facts.
- On our business benchmarks, Ministral 3 8B 2512 ranks higher for agents and tool use, reasoning and knowledge and writing.
At a glance
| Gemma 3 4B | Ministral 3 8B 2512 | |
|---|---|---|
| Price per million tokens | $0.050 in · $0.10 out | $0.15 in · $0.15 out |
| Blended (3 in : 1 out) | $0.062 | $0.15 |
| Cheapest host, blended | – | $0.15 (Mistral) |
| Intelligence Index | 4.8 | 5.5 |
| Spring Prompt overall | – | – |
| Output speed, tokens a second | – | 103 |
| Context window | 131,072 tokens | 262,144 tokens |
| Weights | Open | Open |
| Developer based in | the US | France |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among models available today, for each kind of work, from our own benchmarks and licensed sources. The better rank is in green.
| Use case | Gemma 3 4B | Ministral 3 8B 2512 |
|---|---|---|
| Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls | 160 of 161 | 150 of 161 |
| Reasoning and knowledgeHard questions across science, maths and general knowledge | 187 of 195 | 184 of 195 |
| WritingWhat people prefer in blind comparisons, and judged writing quality | 171 of 187 | 156 of 187 |
| Sticking to the factsSummarising without inventing things, and factual answers | 34 of 153 | 147 of 153 |
Every shared result
| Benchmark | Metric | Gemma 3 4B | Ministral 3 8B 2512 |
|---|---|---|---|
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 2.70 | 9.70 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 4.80 | 5.50 |
| Artificial Analysisreported | GPQA Diamond% of questions | 29.1% | 47.1% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 5.3% | 4.3% |
| Artificial Analysisreported | IFBench% of instructions | 28.3% | 29.1% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 6.7% | 25.7% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 0.4% | 4.1% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 0.8% | 4.5% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 0.4% | 3.7% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 5.0% | 26.6% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 9.0% | 23.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.29 | 0.33 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 24.0 | 28.8 |
| Vectara Hallucination Leaderboardreported | Answer rate% of documents | 67.3% | 99.1% |
| Vectara Hallucination Leaderboardreported | Hallucination rate% of summaries | 6.4% | 21.7% |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.
Quick answers
Is Gemma 3 4B better than Ministral 3 8B 2512?
It depends on the task. Among models available today, on our benchmarks and licensed sources, Gemma 3 4B ranks higher for sticking to the facts; Ministral 3 8B 2512 ranks higher for agents and tool use, reasoning and knowledge and writing.
Which is cheaper, Gemma 3 4B or Ministral 3 8B 2512?
Gemma 3 4B costs $0.062 per million tokens (three input to one output), against $0.15 for Ministral 3 8B 2512.
Run Gemma 3 4B and Ministral 3 8B 2512 on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.