VerdictPal
Join

Models / Google

Gemma 4 31B IT

Gemma · Released 2026-03-01

Google's March 2026 open-weights Gemma 4 instruct model. SAGE 55% on Vals (#2). Small enough to self-host on a single GPU cluster.

Self-hosted Google open weights for privacy-sensitive workflows that can't call cloud APIs.

#36 of 51 on AA Index · current snapshot
29 AA Index · 10 benchmark rows · 1 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite48
Context131K
Input / 1M—
Output / 1M—
Knowledge cutoffopen
Output speed36 t/s
TTFT (default API)0.98s
Statussolid
ReasoningCodingAgenticLong context
Reasoning
86
Coding
43
Agentic
43
Long context
62

59average across 4 tested families

Benchmark placements

Where Gemma 4 31B IT places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
IFBenchReasoning#18/ 5776Artificial Analysis2026-07-21
GPQA DiamondReasoning#40/ 6986Artificial Analysis2026-07-21
Humanity's Last ExamReasoning#44/ 6523Artificial Analysis2026-07-21
AA-LCRLong context#45/ 6462Artificial Analysis2026-07-21
SciCodeCoding#45/ 6543Artificial Analysis2026-07-21
AA time to first tokenPerformance#46/ 69Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#47/ 6415Artificial Analysis2026-07-21
Terminal-BenchAgentic#49/ 6943Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#50/ 6729Artificial Analysis2026-07-21
AA output speedPerformance#58/ 69Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where Gemma 4 31B IT shows up, it's highlighted.

ReasoningGemma 4 31B IT · 86
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingGemma 4 31B IT · 43
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticGemma 4 31B IT · 43
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextGemma 4 31B IT · 62
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Newest receipts

Benchmark rows added to the public ledger for Gemma 4 31B IT in the last 120 days. Older rows live in the full table below.

10 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context62%Artificial Analysis2
AA output speedPerformance36 t/sArtificial Analysis2
AA time to first tokenPerformance0.98sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning29.4%Artificial Analysis2
GPQA DiamondReasoning85.7%Artificial Analysis2
Humanity's Last ExamReasoning22.7%Artificial Analysis2
IFBenchReasoning75.58%Artificial Analysis2
SciCodeCoding43.4%Artificial Analysis2
τ³-Bench BankingAgentic15.05%Artificial Analysis2
Terminal-BenchAgentic43.45%Artificial Analysis2
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Reasoning86
Coding43
Agentic43
Long context62
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for Gemma 4 31B IT. Scores link to the original source; gaps mean no public row exists yet.

10 sourced rows

10 of 32 catalog benchmarks have a sourced row for Gemma 4 31B IT.

Reasoning

Coding

Agentic

Long context

Performance

Sources