Gemma 3 12B
by Google DeepMind · United StatesThe mid-size Gemma 3 — the usual pick for a 24GB card.
Specification
The numbers
- Maker
- Google DeepMind
- Released
- 2025-03
- Parameters
- 12B
- Architecture
- Dense transformer
- Context window
- 131,072 tokens
- Max output
- 8,192 tokens
- Input
- Text, Image
- Output
- Text
- Reasoning
- No
- Tool calling
- Yes
- Knowledge cutoff
- Not reported
- Licence
- Gemma Terms of Use
- Availability
- Not reported
- Weights
- Downloadable
Cost
Price per million tokens
- Input
- $0.05 / M tokens
- Output
- $0.10 / M tokens
- Blended 3:1
- $0.062
This model has open weights, so there is no first-party price. The figures above are a representative third-party hosting rate — you can also run it yourself for the cost of the hardware.
Published scores
Benchmarks
Figures published by Google DeepMind or taken from a public leaderboard. Row last checked 2026-08.
| Category | Score | Rank |
|---|---|---|
| ReasoningGPQA Diamond | 34.9% | 82 of 83 reporting the same tests |
| MathsCompetition mathematics, graded on the final answer | Not reported | — |
| CodingWriting and repairing real code | Not reported | — |
| KnowledgeMMLU-Pro | 60.6% | 39 of 77 reporting the same tests |
| MultimodalMMMU | 59.6% | 29 of 39 reporting the same tests |
| Instruction followingObeying an exact, checkable format | Not reported | — |
| Human preferenceWhich answer people pick, blind | Not reported | — |
A category averages every benchmark in it that Gemma 3 12B reports. The rank counts only models that report the same tests, so it never compares an average over three benchmarks against an average over one.
Every reported test
| MMLU-ProKnowledge | Rank 39 of 77 models reporting | |
|---|---|---|
| GPQA DiamondReasoning | Rank 82 of 83 models reporting | |
| AIME 2025Maths | Not reported | No figure published |
| MATH-500Maths | Not reported | No figure published |
| SWE-bench VerifiedCoding | Not reported | No figure published |
| SWE-bench ProCoding | Not reported | No figure published |
| Terminal-Bench 2.1Coding | Not reported | No figure published |
| Frontier-Bench v0.1Reasoning | Not reported | No figure published |
| Terminal-Bench 4.0Coding | Not reported | No figure published |
| LiveCodeBenchCoding | Not reported | No figure published |
| HumanEvalCoding | Not reported | No figure published |
| MMMUMultimodal | Rank 29 of 39 models reporting | |
| IFEvalInstruction following | Not reported | No figure published |
| LMArena EloHuman preference | Not reported | No figure published |
Where these numbers come from
Every score on this page is a published figure, taken from the model's own card, system card, technical report or release post, or from a public leaderboard. CorX Labs did not run these evaluations. Most are self-reported by the lab that built the model, which means they were produced under that lab's own choice of prompt, scaffold and number of attempts — so treat them as a starting point for a shortlist, not as a settled ranking.
A score someone other than the model's maker measured is marked Independent and names its measurer. Those are the stronger numbers on this page — an outside harness has no reason to flatter anyone — and there are not many of them.
Where a figure has not been published, the cell reads Not reported rather than an estimate. Nothing here is inferred, interpolated or guessed. Each model records the month its row was last checked. Full method and caveats.
Head to head
Gemma 3 12B compared
Same maker
Other models from Google DeepMind
| Gemini 3 ProGoogle DeepMind | Google DeepMind | 1M | $2.00 | $12.00 | — | 91.9% | 95% | 76.2% | Proprietary |
| Gemini 2.5 Flash-LiteGoogle DeepMind | Google DeepMind | 1M | $0.10 | $0.40 | — | 64.6% | 49.8% | — | Proprietary |
| Gemini 2.5 FlashGoogle DeepMind | Google DeepMind | 1M | $0.30 | $2.50 | — | 78.3% | 78% | — | Proprietary |
| Gemini 2.5 ProGoogle DeepMind | Google DeepMind | 1M | $1.25 | $10.00 | — | 84% | 86.7% | 63.8% | Proprietary |
| Gemma 3 27BGoogle DeepMind | Google DeepMind | 131K | $0.10 | $0.20 | 67.5% | 42.4% | — | — | Gemma Terms of Use |
| Gemma 3 4BGoogle DeepMind | Google DeepMind | 131K | $0.02 | $0.04 | 43.6% | — | — | — | Gemma Terms of Use |
| Gemini 2.0 Flash-LiteGoogle DeepMind | Google DeepMind | 1M | $0.075 | $0.30 | 71.6% | 51.5% | — | — | Proprietary |
| Gemini 2.0 FlashGoogle DeepMind | Google DeepMind | 1M | $0.10 | $0.40 | 77.6% | 62.1% | — | — | Proprietary |
| Gemma 2 27BGoogle DeepMind | Google DeepMind | 8.2K | $0.27 | $0.27 | 56% | — | — | — | Gemma Terms of Use |
| Gemma 2 9BGoogle DeepMind | Google DeepMind | 8.2K | $0.06 | $0.06 | 45% | — | — | — | Gemma Terms of Use |
| Gemini 1.5 ProGoogle DeepMind | Google DeepMind | 2.1M | $1.25 | $5.00 | 75.8% | 59.1% | — | — | Proprietary |
| Gemini 1.5 FlashGoogle DeepMind | Google DeepMind | 1M | $0.075 | $0.30 | 67.3% | 51% | — | — | Proprietary |