Google · Open weights
Gemma 4 31B
Google's Gemma 4 31B instruction-tuned model for heavier reasoning, coding, agentic workflows, and long-context multimodal understanding. This route keeps tokenizer thinking disabled for faster direct answers. Requests containing video cost $0.14 per million input tokens and $0.40 per million output tokens.
- Reasoning
- Tools
- Vision
- Structured output
- Video input
- Input
- Text, Images, Video
- Output
- Text
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.14
- Output
- $0.40
- Cache read
- $0.05
Reference prices as published by the model provider. API example
Specifications
- Context window
- 262.1K
- Max output
- 131.1K
- Average output (7 days)
- 338 tokens
- Longer than 36% of 468 models
Side by side
Gemma 4 31B compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| Gemma 4 31BGoogleThis model | DeepSeek V3.1 Terminus (Thinking)DeepSeekRemove | |
|---|---|---|
| Provider | DeepSeek | |
| Discovery tier | Open weights | Open weights |
| Added | Apr 2, 2026 | Jan 1, 2024 |
| Limits | ||
| Context window | 262.14KBest | 128K |
| Max output | 131.07KBest | 65.54K |
| Average output | 338 tokens | 362 tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.14Best | $0.25 |
| Output | $0.40Best | $0.70 |
| Cache read | $0.05 | — |
| Performance | ||
| Throughput | 39.3 tok/sBest | 21.6 tok/s |
| Time to first token | 3.2 sBest | 6.4 s |
| Request success (24h) | 99.7%Best | 99.5% |
| Benchmark indexes | ||
| Intelligence Index | 14.7 | 14.8Best |
| Coding Index | 43.4 | 43.5Best |
| Math Index | — | 89.7 |
| Capabilities | ||
| Reasoning | Yes | No |
| Tools | Yes | Yes |
| Vision | Yes | No |
| Structured output | Yes | Yes |
| Video input | Yes | No |
| Reasoning effort | — | — |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- Gemma 4 31B
- DeepSeek V3.1 Terminus (Thinking)
Intelligence Index
- Gemma 4 31B14.7
- DeepSeek V3.1 Terminus (Thinking)14.8
Gemma 4 31B: Better than 35% of 326 scored models
Indexes
Coding Index
Composite of the coding evaluations
- Gemma 4 31B: 43.4
- DeepSeek V3.1 Terminus (Thinking): 43.5Top
Math Index
Composite of the math evaluations
- Gemma 4 31B: —
- DeepSeek V3.1 Terminus (Thinking): 89.7
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- Gemma 4 31B: 4.9%Top
- DeepSeek V3.1 Terminus (Thinking): 2.7%
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- Gemma 4 31B: 0.5%Top
- DeepSeek V3.1 Terminus (Thinking): 0.2%
AA-Briefcasev1.1
Agentic knowledge work, rated head to head
- Gemma 4 31B: 368 Elo
- DeepSeek V3.1 Terminus (Thinking): 458 EloTop
GDPval-AAv2.1
Economically valuable occupational tasks, rated head to head
- Gemma 4 31B: 619 Elo
- DeepSeek V3.1 Terminus (Thinking): 709 EloTop
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- Gemma 4 31B: 6.0%Top
- DeepSeek V3.1 Terminus (Thinking): 5.4%
AA-LCRv1.1
Reasoning across long documents
- Gemma 4 31B: 69.7%Top
- DeepSeek V3.1 Terminus (Thinking): 69.3%
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- Gemma 4 31B: 23.6%Top
- DeepSeek V3.1 Terminus (Thinking): 16.4%
GPQA Diamond
Graduate-level science questions
- Gemma 4 31B: 85.7%Top
- DeepSeek V3.1 Terminus (Thinking): 79.2%
MMLU-Pro
Multi-subject knowledge and reasoning
- Gemma 4 31B: —
- DeepSeek V3.1 Terminus (Thinking): 85.1%
IFBench
Following precise output instructions
- Gemma 4 31B: 75.6%Top
- DeepSeek V3.1 Terminus (Thinking): 57.0%
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- Gemma 4 31B: 0.0%Top
- DeepSeek V3.1 Terminus (Thinking): 0.0%Top
SciCode
Scientific computing in Python
- Gemma 4 31B: 45.5%Top
- DeepSeek V3.1 Terminus (Thinking): 38.0%
LiveCodeBench
Recent competitive programming problems
- Gemma 4 31B: —
- DeepSeek V3.1 Terminus (Thinking): 79.8%
Math
AIME 2025
Competition mathematics, 2025 problems
- Gemma 4 31B: —
- DeepSeek V3.1 Terminus (Thinking): 89.7%
Legacy evaluations
Terminal-Bench Hard
Earlier terminal task suite
- Gemma 4 31B: 36.4%Top
- DeepSeek V3.1 Terminus (Thinking): 30.3%
τ²-Bench Telecom
Tool use in a customer-support simulation
- Gemma 4 31B: 59.9%Top
- DeepSeek V3.1 Terminus (Thinking): 37.1%
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- Gemma 4 31B: 69.7%Top
- DeepSeek V3.1 Terminus (Thinking): 69.3%
Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis
Live routing data
Performance and reliability
Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.
- Throughput
- 39.3 tok/s
- Faster than 27% of 254 models
- Time to first token
- 3.2 s
- Quicker than 55% of 254 models
- Prompt cache hits
- 15.1%
- About $0.13 per 1M input tokens with caching
- Request success (24h)
- 99.7%
- Reliable
Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.
- Healthy (98% or more)
- Degraded (90–98%)
- Down (under 90%)
- No requests
Developers
Use it through the API
Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
curl https://www.noumen.xyz/api/v1/chat/completions \
-H "Authorization: Bearer $NOUMEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google/gemma-4-31b-it",
"messages": [{ "role": "user", "content": "Explain this stack trace." }]
}'import os
from openai import OpenAI
client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
model="google/gemma-4-31b-it",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)Related models
Other Google models, starting with the same family.
