Gemma 4 31B

google/gemma-4-31b-it-20260402

Gemma 4 31B is a Google model that accepts image, text, and video inputs and generates text output, with a context length of 262 144 tokens. It supports tools and reasoning, is released with open weights as of 2026‑04‑02, and charges $0.12 per million input tokens and $0.37 per million output tokens.

  • tool use
  • structured outputs
  • reasoning
  • open-weight

At a glance

providergoogle
releasedApr 2, 2026
context262K
input / M$0.090

Specifications

What the API gives you

Context window
262K tokens
Max output
16K tokens
Input price
$0.090 / 1M tokens
Output price
$0.34 / 1M tokens
Cache read
$0.050 / 1M tokens
Modalities
image, text, video → text
Knowledge cutoff
Released
Apr 2, 2026

Measured evidence

Public benchmark record

LMArena Agent
-13.62026-07-13
LMArena Agent
-142026-07-20
LMArena Agent
-15.92026-07-27
LMArena Agent
-14.52026-07-21
LMArena Agent
-16.12026-07-28
LMArena Agent
-17.22026-08-03
LMArena Agent
-18.42026-08-12
LMArena Agent
-17.62026-08-06
LMArena Agent
-18.22026-08-11
LMArena Agent
-18.92026-08-13
LMArena Agent
-20.62026-08-24
LMArena Agent
-22.32026-08-31
LMArena Agent
-23.72026-09-05
LMArena Agent
-21.32026-08-27
LMArena Creative Writing
1417 Elo2026-07-21
LMArena Creative Writing
1417 Elo2026-07-27
LMArena Creative Writing
1417 Elo2026-07-30
LMArena Creative Writing
1417 Elo2026-08-02
LMArena Creative Writing
1417 Elo2026-08-03
LMArena Creative Writing
1416 Elo2026-08-06
LMArena Creative Writing
1418 Elo2026-08-12
LMArena Creative Writing
1417 Elo2026-08-21
LMArena Creative Writing
1416 Elo2026-09-02
LMArena Creative Writing
1416 Elo2026-08-27
LMArena Creative Writing
1416 Elo2026-09-01
LMArena Long Query
1466 Elo2026-07-21
LMArena Long Query
1443 Elo2026-07-27
LMArena Long Query
1443 Elo2026-07-30
LMArena Long Query
1443 Elo2026-08-02
LMArena Long Query
1443 Elo2026-08-03
LMArena Long Query
1443 Elo2026-08-06
LMArena Long Query
1444 Elo2026-08-12
LMArena Long Query
1444 Elo2026-08-21
LMArena Long Query
1444 Elo2026-09-02
LMArena Long Query
1443 Elo2026-08-27
LMArena Long Query
1443 Elo2026-09-01
LMArena Math
1465 Elo2026-07-21
LMArena Math
1466 Elo2026-07-27
LMArena Math
1466 Elo2026-07-30
LMArena Math
1466 Elo2026-08-02
LMArena Math
1466 Elo2026-08-03
LMArena Math
1466 Elo2026-08-06
LMArena Math
1467 Elo2026-08-12
LMArena Math
1468 Elo2026-08-21
LMArena Math
1468 Elo2026-09-01
LMArena Math
1468 Elo2026-08-27
LMArena Text
1442 Elo2026-07-16
LMArena Text
1451 Elo2026-08-03
LMArena Text
1451 Elo2026-07-21
LMArena Text
1451 Elo2026-08-02
LMArena Text
1451 Elo2026-07-27
LMArena Text
1451 Elo2026-07-30
LMArena Text
1451 Elo2026-08-10
LMArena Text
1451 Elo2026-08-06
LMArena Text
1451 Elo2026-08-21
LMArena Text
1451 Elo2026-08-12
LMArena Text
1451 Elo2026-08-27
LMArena Text
1451 Elo2026-09-01
LMArena Text
1451 Elo2026-09-02
LMArena Vision
1270 Elo2026-07-12
LMArena Vision
1256 Elo2026-07-21
LMArena Vision
1257 Elo2026-07-26
LMArena Vision
1257 Elo2026-07-30
LMArena Vision
1257 Elo2026-08-02
LMArena Vision
1256 Elo2026-08-06
LMArena Vision
1260 Elo2026-08-21
LMArena Vision
1261 Elo2026-08-27
LMArena WebDev
1367 Elo2026-07-16
LMArena WebDev
1367 Elo2026-07-21
LMArena WebDev
1367 Elo2026-07-20
LMArena WebDev
1364 Elo2026-07-27
LMArena WebDev
1364 Elo2026-07-24
LMArena WebDev
1364 Elo2026-07-28
LMArena WebDev
1364 Elo2026-07-30
LMArena WebDev
1363 Elo2026-08-01
LMArena WebDev
1363 Elo2026-07-31
LMArena WebDev
1364 Elo2026-08-06
LMArena WebDev
1363 Elo2026-08-02
LMArena WebDev
1365 Elo2026-08-12
LMArena WebDev
1366 Elo2026-08-10
LMArena WebDev
1364 Elo2026-08-14
LMArena WebDev
1363 Elo2026-09-01
LMArena WebDev
1363 Elo2026-09-05
LMArena WebDev
1364 Elo2026-08-21
LMArena WebDev
1362 Elo2026-08-30
SWE-rebench
25.1% ± 1.02026-05-15tools

LMArena data licensed CC-BY-4.0. Aider (Apache-2.0) and SWE-bench (MIT) are open. Design Arena uses blind human preference. Submitted scaffold plus model result. Compare the harness and effort settings, not only the model name.

Open-weight model on Hugging Face.

Task evidence

How broad is the support?

Each status is exact to this model version and the selected task. Official release claims remain provisional.

  • Best LLM for Business WritingProvisional evidence

    70% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Marketing CopyProvisional evidence

    70% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Chat & RoleplayProvisional evidence

    70% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 5 of 5 configured evaluation feeds currently have results.

  • 70% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Creative WritingProvisional evidence

    60% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for WritingProvisional evidence

    60% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 5 of 5 configured evaluation feeds currently have results.

  • 55% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 6 of 6 configured evaluation feeds currently have results.

  • 55% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Reasoning or VLMEvalKit tasks. 3 of 4 configured evaluation feeds currently have results.

  • 55% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Reasoning. 5 of 5 configured evaluation feeds currently have results.

  • 50% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Reasoning. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Massive ContextProvisional evidence

    50% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LongBench v2. 3 of 4 configured evaluation feeds currently have results.

  • Best LLM for OCR & DocumentsProvisional evidence

    45% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Structured-output eval or VLMEvalKit tasks. 3 of 5 configured evaluation feeds currently have results.

  • 45% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Structured-output eval or VLMEvalKit tasks. 3 of 5 configured evaluation feeds currently have results.

  • Best LLM for Charts & DiagramsProvisional evidence

    45% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Structured-output eval or VLMEvalKit tasks. 3 of 5 configured evaluation feeds currently have results.

  • Best LLM for Document ParsingProvisional evidence

    45% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Structured-output eval or VLMEvalKit tasks. 3 of 5 configured evaluation feeds currently have results.

  • Best LLM for Hermes AgentProvisional evidence

    45% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Berkeley Function Calling or Terminal-Bench 2.1. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for OpenHandsProvisional evidence

    45% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Berkeley Function Calling or Terminal-Bench 2.1. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for GooseProvisional evidence

    45% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Berkeley Function Calling or Terminal-Bench 2.1. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for OpenClawProvisional evidence

    45% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Berkeley Function Calling or Terminal-Bench 2.1. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for CodingProvisional evidence

    44% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Aider Polyglot or Design Arena Coding. 8 of 8 configured evaluation feeds currently have results.

  • Best LLM for Code CompletionProvisional evidence

    44% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 4 of 4 configured evaluation feeds currently have results.

  • 40% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from FACTS Grounding or LiveBench Instruction Following. 4 of 6 configured evaluation feeds currently have results.

  • Best LLM for Competition MathProvisional evidence

    40% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Mathematics or LiveBench Reasoning. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for SummarizationProvisional evidence

    40% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from FACTS Grounding or LiveBench Instruction Following. 4 of 6 configured evaluation feeds currently have results.

  • Best LLM for Math & ReasoningProvisional evidence

    40% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Mathematics or LiveBench Reasoning. 5 of 5 configured evaluation feeds currently have results.

  • Best Local LLM for CodingProvisional evidence

    40% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Autonomous AgentsProvisional evidence

    37% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Frontier-Bench or LiveBench Agentic Coding. 7 of 7 configured evaluation feeds currently have results.

  • Best Value LLMProvisional evidence

    36% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or Route reliability. 3 of 4 configured evaluation feeds currently have results.

  • 35% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from FACTS Grounding or LMArena Search. 4 of 6 configured evaluation feeds currently have results.

  • 33% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or Route reliability. 3 of 4 configured evaluation feeds currently have results.

  • Best LLM for Agentic CodingProvisional evidence

    32% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Design Arena Full-stack or Design Arena Web-apps. 8 of 8 configured evaluation feeds currently have results.

  • Best LLM for AI AgentsProvisional evidence

    32% of intended weighted task evidence across 2 independent sources; broader independent confirmation is still missing from Berkeley Function Calling or LiveBench Agentic Coding. 7 of 7 configured evaluation feeds currently have results.

  • Best LLM for Frontend & UIProvisional evidence

    30% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Design Arena Coding or Design Arena UI. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Fast SummariesProvisional evidence

    29% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or Route reliability. 3 of 4 configured evaluation feeds currently have results.

  • 25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 4 of 5 configured evaluation feeds currently have results.

  • Best Cheap LLM for High VolumeProvisional evidence

    25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or Route reliability. 3 of 4 configured evaluation feeds currently have results.

  • 25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 4 of 5 configured evaluation feeds currently have results.

  • Cheapest Capable LLMProvisional evidence

    25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or Route reliability. 3 of 4 configured evaluation feeds currently have results.

  • Best LLM for TranslationProvisional evidence

    25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Instruction Following or LiveBench Language. 4 of 5 configured evaluation feeds currently have results.

  • 25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or LMArena Search. 4 of 5 configured evaluation feeds currently have results.

  • 25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or OSWorld Verified. 4 of 4 configured evaluation feeds currently have results.

  • 17% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or LiveBench Data Analysis. 4 of 5 configured evaluation feeds currently have results.

  • 17% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or LiveBench Data Analysis. 4 of 5 configured evaluation feeds currently have results.

  • 13% of intended weighted task evidence across 0 independent sources; broader independent confirmation is still missing from LiveBench Data Analysis or Route reliability. 2 of 4 configured evaluation feeds currently have results.

  • Best LLM for Grounded RAGProvisional evidence

    10% of intended weighted task evidence across 0 independent sources; broader independent confirmation is still missing from CRAG or FACTS Grounding. 2 of 5 configured evaluation feeds currently have results.

Recommendation coverage

Where this model appears

Not currently included in a published recommendation. The task evidence panel above says exactly what is confirmed and what is still missing.

Catalog data via OpenRouter and models.dev. Last checked within the past 6 hours.