Google · Open weights
Gemma 4 31B Scarlet
Scarlet is a multimodal Gemma 4 31B creative finetune for expressive dialogue, long-form storytelling, and roleplay.
- Reasoning
- Tools
- Vision
- Input
- Text, Images
- Output
- Text
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.10
- Output
- $0.45
- Cache read
- $0.05
Reference prices as published by the model provider. API example
Specifications
- Context window
- 262.1K
- Max output
- 32.8K
- Average output (7 days)
- 456 tokens
- Longer than 45% of 473 models
Side by side
Gemma 4 31B Scarlet compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| Gemma 4 31B ScarletGoogleThis model | Gemini 4 ArgonGoogleRemove | |
|---|---|---|
| Provider | ||
| Discovery tier | Open weights | Frontier |
| Added | Oct 8, 2026 | Oct 13, 2026 |
| Limits | ||
| Context window | 262.14K | 1.05MBest |
| Max output | 32.77K | 65.54KBest |
| Average output | 456 tokens | 1.6K tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.10Best | $2.00 |
| Output | $0.45Best | $10.00 |
| Cache read | $0.05Best | $0.10 |
| Cache write | — | $0.19 |
| Performance | ||
| Throughput | 15.5 tok/s | 66.1 tok/sBest |
| Time to first token | 8.5 s | 6.9 sBest |
| Request success (24h) | 94.1% | 98.5%Best |
| Benchmark indexes | ||
| Intelligence Index | — | 52.6 |
| Capabilities | ||
| Reasoning | Yes | Yes |
| Tools | Yes | Yes |
| Vision | Yes | Yes |
| Structured output | No | Yes |
| PDF input | No | Yes |
| Audio input | No | Yes |
| Video input | No | Yes |
| Reasoning effort | — | low, medium, high |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- Gemma 4 31B Scarlet
- Gemini 4 Argon
Intelligence Index
- Gemma 4 31B Scarlet—
- Gemini 4 Argon52.6
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 77.5%
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 53.1%
Harvey LAB-AAv1.1
Legal tasks passed in full, with no material hallucination
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 3.9%
AA-Briefcasev1.1
Agentic knowledge work, rated head to head
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 1,486 Elo
GDPval-AAv2.1
Economically valuable occupational tasks, rated head to head
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 1,624 Elo
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 21.8%
AA-LCRv1.1
Reasoning across long documents
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 79.7%
MLCR-AA
Long-context reasoning over medical records
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 13.9%
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 57.1%
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 57.1%
SciCode
Scientific computing in Python
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 61.8%
Legacy evaluations
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- Gemma 4 31B Scarlet: —
- Gemini 4 Argon: 79.7%
Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Combined | Not in top 20 | #2 · 55.6% |
Ranks among Noumen catalog models on Combined leaderboards. Only top-20 placements are listed.
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Roleplay | Not in top 20 | #16 · 22.0% |
| Creative writing | Not in top 20 | #7 · 41.3% |
| Coding | Not in top 20 | #7 · 30.0% |
| Math | Not in top 20 | #1 · 50.0% |
Ranks among Noumen catalog models on Use cases leaderboards. Only top-20 placements are listed.
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Overall | Not in top 20 | #1 · 1517.2 |
| Coding | Not in top 20 | #1 · 1551.6 |
| Math | Not in top 20 | #1 · 1531.7 |
| Longer Query | Not in top 20 | #1 · 1534.5 |
| Creative Writing | Not in top 20 | #2 · 1512.4 |
| Instruction Following | Not in top 20 | #2 · 1519.2 |
| Hard Prompts | Not in top 20 | #1 · 1540.3 |
| Chinese | Not in top 20 | #1 · 1579.2 |
| English | Not in top 20 | #1 · 1520.1 |
| Exclude Ties | Not in top 20 | #1 · 1542.6 |
| Expert | Not in top 20 | #5 · 1537.0 |
| French | Not in top 20 | #1 · 1541.0 |
| Hard Prompts English | Not in top 20 | #1 · 1537.8 |
| Industry Business And Management And Financial Operations | Not in top 20 | #1 · 1508.8 |
| Industry Entertainment And Sports And Media | Not in top 20 | #1 · 1508.5 |
| Industry Legal And Government | Not in top 20 | #2 · 1514.5 |
| Industry Life And Physical And Social Science | Not in top 20 | #1 · 1531.9 |
| Industry Mathematical | Not in top 20 | #1 · 1539.6 |
| Industry Medicine And Healthcare | Not in top 20 | #1 · 1533.3 |
| Industry Software And It Services | Not in top 20 | #1 · 1548.9 |
| Industry Writing And Literature And Language | Not in top 20 | #2 · 1515.0 |
| Multi Turn | Not in top 20 | #1 · 1543.8 |
| Non English | Not in top 20 | #1 · 1508.0 |
| Polish | Not in top 20 | #2 · 1516.4 |
| Russian | Not in top 20 | #2 · 1528.8 |
| Spanish | Not in top 20 | #5 · 1498.2 |
| WebDev Overall | Not in top 20 | #9 · 1674.2 |
| WebDev · Webdev | Not in top 20 | #9 · 1674.2 |
| WebDev · Webdev-html | Not in top 20 | #13 · 1641.7 |
| WebDev · Webdev-react | Not in top 20 | #8 · 1689.8 |
Ranks among Noumen catalog models on LMArena leaderboards. Only top-20 placements are listed.LMArena
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| FrontierCode Main | Not in top 20 | #11 · 47.5% |
| FrontierCode Extended | Not in top 20 | #9 · 60.6% |
Ranks among Noumen catalog models on FrontierCode leaderboards. Only top-20 placements are listed.FrontierCode
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Epoch Capabilities Index | Not in top 20 | #5 · 166.4 |
Ranks among Noumen catalog models on Epoch AI leaderboards. Only top-20 placements are listed.Epoch AI
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| ARC-AGI-1 | Not in top 20 | #4 · 98.5% |
| ARC-AGI-2 | Not in top 20 | #5 · 91.3% |
| ARC-AGI-3 (Default harness) | Not in top 20 | #3 · 49.8% |
Ranks among Noumen catalog models on ARC Prize leaderboards. Only top-20 placements are listed.ARC Prize
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| EQ-Bench 4 | Not in top 20 | #11 · 1240 |
| Longform Writing | Not in top 20 | #14 · 80.9 |
Ranks among Noumen catalog models on EQ-Bench leaderboards. Only top-20 placements are listed.EQ-Bench
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Game Development | Not in top 20 | #8 · 1370 |
| Website | Not in top 20 | #9 · 1317 |
| UI Component | Not in top 20 | #11 · 1332 |
| Data Visualization | Not in top 20 | #8 · 1332 |
| 3D | Not in top 20 | #9 · 1384 |
Ranks among Noumen catalog models on Design Arena leaderboards. Only top-20 placements are listed.Design Arena
- Gemma 4 31B Scarlet
- Gemini 4 Argon
| Leaderboard | Gemma 4 31B Scarlet | Gemini 4 Argon |
|---|---|---|
| Usage Share | Not in top 20 | #3 · 8.8% |
Ranks among Noumen catalog models on Usage leaderboards. Only top-20 placements are listed.
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
- 15.5 tok/s
- Faster than 5% of 260 models
- Time to first token
- 8.5 s
- Quicker than 13% of 260 models
- Prompt cache hits
- 6.0%
- About $0.097 per 1M input tokens with caching
- Request success (24h)
- 94.1%
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.
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": "gemma-4-31b-it-scarlet",
"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="gemma-4-31b-it-scarlet",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)Related models
Other Google models, starting with the same family.