Google model product
Gemma 4 31B
Gemma 4 31B is Google DeepMind's flagship dense multimodal model with 31 billion parameters and a 256K context window.
Updated Aug 12, 2026. Default version: Gemma 4 31B
LLMBoard score60.1Gemma 4 31B
Coverage80%12 benchmark families
Context window262.1KTokens
Official input priceN/AOfficial price unavailable
Gemma 4 31B Specifications
Technical details for the model's default version.
- Version
- Gemma 4 31B
- Released
- Apr 2, 2026
- Knowledge cutoff
- Jan 1, 2025
- Parameters
- 30.7B
- Context window
- 262.1K
- Max output
- 32.8K
- Inputs
- image, text
- Outputs
- text
- Open weights
- Yes
- License
- Apache 2.0
Gemma 4 31B Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Gemma 4 31B Benchmark Results
Benchmark scores for Gemma 4 31B.
| BIG-Bench Extra Hard | 74.4% | 01 | 11 | 100.0% | C | |
| MedXpertQA | 61.3% | 05 | 12 | 63.6% | C | |
| MRCR v2 (8-needle) | 66.4% | 05 | 21 | 80.0% | C | |
| t2-bench | 86.4% | 05 | 23 | 81.8% | C | |
| MathVision | 85.6% | 09 | 32 | 74.2% | C | |
| AIME 2026 | 89.2% | 12 | 18 | 35.3% | C | |
| MMLU-Pro | 85.2% | 15 | 129 | 89.1% | C | |
| MMMLU | 88.4% | 18 | 49 | 64.6% | C | |
| LiveCodeBench v6 | 80.0% | 21 | 53 | 61.5% | C | |
| MMMU-Pro | 76.9% | 24 | 66 | 64.6% | C | |
| Humanity's Last Exam | 26.5% | 47 | 93 | 50.0% | C | |
| GPQA | 84.3% | 59 | 234 | 75.1% | C |
Gemma 4 31B Arena Results
Preference and agent-evaluation results for the default version.
| agent praise complaint | overall | 28 | -0.0 | N/A | 10.6K | |
| document style control | overall | 28 | 1445.2 | 10,577 | N/A | |
| agent task outcome explicit | overall | 29 | -0.0 | N/A | 34.8K | |
| document | overall | 29 | 1423.7 | 10,577 | N/A | |
| vision | overall | 31 | 1270.1 | 32,461 | N/A | |
| vision style control | overall | 35 | 1255.7 | 32,461 | N/A | |
| agent steerability | overall | 40 | -0.1 | N/A | 42.8K | |
| agent | overall | 46 | -0.2 | N/A | 707.6K | |
| agent bash recovery steps | overall | 46 | -0.4 | N/A | 27.1K | |
| agent tool hallucination | overall | 46 | -0.3 | N/A | 592.3K | |
| text factuality | overall | 53 | 1449.0 | 5,898 | N/A | |
| text | overall | 56 | 1441.5 | 5,901 | N/A | |
| text style control | overall | 58 | 1451.2 | 5,901 | N/A | |
| webdev | overall | 73 | 1365.6 | 8,349 | N/A |
Gemma 4 31B Runtime Performance
Provider-specific output speed and catalog latency for Gemma 4 31B. Runtime does not affect the capability score.
| FriendliAI | 20.155 tok/s | 1.717 s | 262.1K | 131.1K | |
| Novita | 1.346 tok/s | 2.027 s | 262.1K | 131.1K | |
| DeepInfra | 0.94 tok/s | 37.139 s | 262.1K | 131.1K | |
| Together | 0.852 tok/s | N/A | 262.1K | 131.1K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Gemma 4 31B Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.08 input, $0.35 output per 1M via Kilo Gateway
- Tracked offerings
- 32
| Poe | google/gemma-4-31b | global | N/A | N/A | 262.1K | |
| Nvidia | google/gemma-4-31b-it | global | N/A | N/A | 256K | |
| Kenari | gemma-4-31b-it | global | N/A | N/A | 262.1K | |
| Kilo Gateway | google/gemma-4-31b-it | global | $0.08 | $0.35 | 262.1K | |
| Weights & Biases | google/gemma-4-31B-it | global | $0.10 | $0.34 | 262.1K | |
| NanoGPT | google/gemma-4-31b-it | global | $0.10 | $0.35 | 262.1K | |
| CrofAI | gemma-4-31b-it | global | $0.10 | $0.30 | 262.1K | |
| OpenRouter | google/gemma-4-31b-it | global | $0.10 | $0.34 | 262.1K | |
| LLM Gateway | gemma-4-31b-it | global | $0.102 | $0.297 | 262.1K | |
| Lilac | google/gemma-4-31b-it | global | $0.11 | $0.35 | 262.1K | |
| FastRouter | google/gemma-4-31b-it | global | $0.13 | $0.38 | 262.1K | |
| NEAR AI Cloud | google/gemma-4-31B-it | global | $0.13 | $0.40 | 262.1K | |
| SiliconFlow | google/gemma-4-31B-it | global | $0.13 | $0.40 | 262.1K | |
| OrcaRouter | google/gemma-4-31b-it | global | $0.13 | $0.38 | 262.1K | |
| Deep Infra | google/gemma-4-31B-it | global | $0.13 | $0.38 | 262.1K | |
| Friendli | google/gemma-4-31B-it | global | $0.14 | $0.40 | 262.1K | |
| Abacus | google/gemma-4-31b-it | global | $0.14 | $0.40 | 262.1K | |
| Merge Gateway | google/gemma-4-31b-it | global | $0.14 | $0.40 | 262.1K | |
| Vercel AI Gateway | google/gemma-4-31b-it | global | $0.14 | $0.40 | 256K | |
| NovitaAI | google/gemma-4-31b-it | global | $0.14 | $0.40 | 262.1K | |
| Hugging Face | google/gemma-4-31B-it | global | $0.14 | $0.40 | 262.1K | |
| Neuralwatt | gemma-4-31b | global | $0.144 | $0.42 | 262.1K | |
| DigitalOcean | gemma-4-31B-it | global | $0.18 | $0.50 | 256K | |
| ai& | google/gemma-4-31b-it | global | $0.20 | $0.50 | 262.1K | |
| Cortecs | gemma-4-31b-it | global | $0.223 | $0.39 | 262K | |
| Infomaniak | google/gemma-4-31B-it | global | $0.25 | $0.50 | 100K | |
| Berget.AI | google/gemma-4-31B-it | global | $0.275 | $0.55 | 128K | |
| Together AI | google/gemma-4-31B-it | global | $0.39 | $0.97 | 262.1K | |
| NanoGPT | TEE/gemma4-31b | global | $0.45 | $1 | 262.1K | |
| Pioneer | google/gemma-4-31B-it | global | $0.50 | $0.50 | 32.8K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Gemma 4 31B Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Gemma 4 31B | 60.1 | 30.7B | 262.1K | 32.8K | Yes | Apache 2.0 |
What is Gemma 4 31B?
Key information about Gemma 4 31B and its available data.
Gemma 4 31B is Google DeepMind's flagship dense multimodal model with 31 billion parameters and a 256K context window. Ranks #3 among open models on Arena AI.
Built from the same research as Gemini 3, it features Per-Layer Embeddings, Shared KV Cache, alternating sliding-window and global attention, and variable aspect ratio vision encoding. Achieves an estimated LMArena text score of 1452.
Data as of 2026-08-11.
Gemma 4 31B vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Gemma 4 31B
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
GO
Google
Gemini 3.5 Flash Lite
GOGoogle
Gemini 3 Flash
GOGoogle
Gemini 3 Pro
GOGoogle
Gemma 4 26B A4B
GOGoogle
Gemini 3.1 Flash Lite
GOGoogle
Gemini 3.5 Flash Cyber
FAQ
Common questions about Gemma 4 31B.
When was Gemma 4 31B released?
Gemma 4 31B's default version was released on Apr 2, 2026.
How much does Gemma 4 31B cost?
No official standard PAYG price is currently available for Gemma 4 31B. The lowest tracked third-party offer starts at $0.08 input and $0.35 output via Kilo Gateway.
Who created Gemma 4 31B?
Gemma 4 31B was created by Google.
What is the context window for Gemma 4 31B?
The default version has a 262.1K token context window.
Is Gemma 4 31B open weight?
Yes. The default version is marked as open weight under Apache 2.0.
How many API providers offer Gemma 4 31B?
32 provider offerings are linked to the default version.
What models should I compare Gemma 4 31B with?
Nearby ranked alternatives include Qwen3.6 27B, GPT-5.4-mini, Step 3.5 Flash.