Google · apache_2_0

Gemma 4 31B

Gemma 4 31B by Google appears in 2 sources with Reasoning at 44.8. Best read for Coding, High intelligence, Low cost.

CreatorGoogle

Release date2026-04-02

Knowledge cutoffNot published

Context262K tokens

Input price$0.14/M tokens

Output price$0.4/M tokens

Modalitytext + vision

CountryUS

Metrics

All source-backed metrics

Open in leaderboard

LLM Stats Rank35ranking · source_rank

LLM Stats Code Index (estimated from arena)28.78 scoremetric

Reasoning44.8reasoning · index_reasoning

Math39.45math · index_math

Vision27.1multimodal · index_vision

Tool calling19.41tool_calling · index_tool_calling

Long context29.85long_context · index_long_context

Finance41.91domain · index_finance

Legal41.89domain · index_legal

Healthcare36.32domain · index_healthcare

GPQA84.3 %reasoning · gpqa_score

Code Arena1,135.37 %coding · coding_arena_score

Humanity Last Exam26.5 %reasoning · hle_score

MMMLU88.4 %reasoning · mmmlu_score

MMMU-Pro76.9 %multimodal · mmmu_pro_score

Context262,144 tokenscontext · context

Speed211.34 c/sperformance · throughput

Latency465.35 msperformance · latency

Input price0.14 $/Mpricing · input_price

Output price0.4 $/Mpricing · output_price

Parameters30,700,000,000 paramsmodel · params

Arena Rating1,441.98metric

Arena Rank35metric

Vote Count5,829metric

Evidence

Citations and source overlap

Arena AILM Arena text leaderboard (overall) mirrored in the public leaderboard dataset.Retrieved 2026-05-20 LLM StatsPublic full LLM leaderboard advanced explorer snapshot, including ranking, benchmark indexes, pricing, speed, context, license, and model metadata.Retrieved 2026-06-01

FAQ

How should I read this profile?

Treat this as a source-backed model dossier, not an EvalKit-run verification. The public values are replicated from linked sources and kept source-scoped.

Is Gemma 4 31B verified by EvalKit?

No. EvalKit currently shows 0 verified rows until real run evidence exists.

Why can metrics disagree?

Different sources test different tasks, dates, prompts, and aggregation methods. EvalKit keeps those differences visible instead of merging them into a fake universal score.