GoogleGemma 3 12B
Gemma 3 12B is a Google open-weights model, released 12 Mar 2025. Its best configuration ranks #741 of 785 on the BenchLeader Index at 36.9 ±5.2, in the lower half. The ± is the point: 193 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in human preference (50) and lowest in coding (25). At $0.075 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 45 tokens per second puts it in the slowest quarter, with a first token in 0.7 s. Last measured 22 Oct 2026.
- Blended price
- $0.075/M
- $0.050 in · $0.150 out
- Output speed
- 45 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 0.65 s
- Context
- 131k
- Full answer
- –
- Cost per run
- –
- Released
- 12 Mar 2025
- Overall index37
- Reasoning38
- Coding25
- Agents & tools35
- Maths43
- Knowledge27
- Instruction following40
- Human preference50
- Multimodal27
- Long context28
- Composite34
Versions
Google has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Gemma 4 12Bthinking | 3 Jun 2026 | 50.5 | #370 |
| Gemma 3 12Bthis page | 12 Mar 2025 | 36.9 | #741 |
How its index has moved
8 Sept 2026 to 23 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| GPQA Diamond | 39.5%#250 | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | 1332#230 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 35.0%#494 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 4.2%#525 | Artificial Analysis | |
| CritPt | 0.0%#315 | Artificial Analysis | |
| LMCAnot in index | 4.5%#191 | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | 48.8%#217 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| SciCode | 16.4%#213 | SciCode | |
| Arena Coding | 1316#264 | Arena (LMArena) | |
| SciCode (AA)not in index | 16.4%#228 | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | 0.0%#152 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 0.0%#232 | Artificial Analysis |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 0.8%#349 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 10.8%#380 | Artificial Analysis | |
| BFCL Overall | 30.4%#49 | Berkeley Function Calling Leaderboard | |
| τ³-Banking (AA) | 0.8%#201 | Artificial Analysis | |
| AutomationBenchnot in index | 0.2%#219 | Artificial Analysis | |
| GDP.pdf (AA)not in index | 1.0%#204 | Artificial Analysis | |
| BrowseCompnot in index | 0.5%#29 | OpenAI |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| OTIS Mock AIME | 16.7%#224 | Epoch AI Benchmarking Hub | |
| Arena Maths | 1318#215 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -77.1#531 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 10.3%#493 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 2.6%#546 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 36.7%#296 | Artificial Analysis | |
| Arena Instruction Followingnot in index | 1321#215 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1342#212 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1333#186 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1344#201 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1347#204 | Arena (LMArena) |
Multimodal
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| MMMU-Pro | 37.5%#270 | Artificial Analysis | |
| MMMU (validation) | 59.6%#33 | MMMU | |
| MindCubenot in index | 46.7%#1 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 8.3%#473 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 123.5#165 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 3.8#632 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| NextBit | 59 tok/s | 1.23 s | $0.075 | $0.150 | 131k | int4 |
| DeepInfra | 33 tok/s | 2.92 s | $0.040 | $0.120 | 131k | bf16 |
| DeepInfra | 23 tok/s | 0.99 s | $0.050 | $0.150 | 131k | bf16 |
| DeepInfra | 23 tok/s | 1.86 s | $0.075 | $0.225 | 131k | bf16 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0001 | – | 7.3 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0007 | – | 14.0 s |
| Code edit | 6,000 / 1,500 | $0.0005 | – | 34.0 s |
| Agentic coding session | 60,000 / 4,000 | $0.0036 | – | 1.5 min |
| Structured extraction | 2,000 / 200 | $0.0001 | – | 5.1 s |
More from Google
See also
Questions people ask
- What are Gemma 3 12B's benchmark scores?
- As of 25 Oct 2026, Gemma 3 12B scores 36.9 ±5.2 on the BenchLeader Index, strongest in human preference (50) and weakest in coding (25). That rests on 17 independent benchmark results, and every score on this page links to its publisher.
- How does Gemma 3 12B rank?
- Gemma 3 12B is #741 of 785 ranked configurations on 25 Oct 2026. 193 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Gemma 3 12B cost per 1M tokens?
- As of 25 Oct 2026, Gemma 3 12B costs $0.050 per 1M input tokens and $0.150 per 1M output tokens, $0.075 blended (three input tokens per output token). The list price last changed on 11 Sept 2026.
- When was Gemma 3 12B released?
- Google released Gemma 3 12B on 12 Mar 2025.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Gemma 3 12B: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/gemma-3-12b.