ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

Gemma 4 26B A4BvsGemma 4 31B

Gemma 4 31B leads 6 of 6 shared benchmarks. Both offer a 256K-token context window.

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
0 – 6
Gemma 4 31B leads
Cheaper per token
list price, input + output
Larger context
Tie
both 256K tokens
Provider uptime (30d)
100%
Google
Gemma 4 26B A4B
Google · released 2026-04-02
textvision
Context
256K
Max out
Input /1M
Output /1M
Cached /1M
Scores
6 · 5 core
Gemma 4 31B
Google · released 2026-04-02
textvision
Context
256K
Max out
Input /1M
Output /1M
Cached /1M
Scores
6 · 5 core
Quality

Benchmark matrix

BenchmarkGemma 4 26B A4BGemma 4 31BΔEdge
Reported by both · 6
GPQA Diamond82.3% 84.3% -2 ptGemma 4 31B
MMLU-Pro82.6% 85.2% -2.6 ptGemma 4 31B
AIME 202688.3% 89.2% -0.9 ptGemma 4 31B
Humanity's Last Exam (no tools)8.7% 19.5% -10.8 ptGemma 4 31B
LMArena Elo1434.5 1441.7 -7.2Gemma 4 31B
LiveCodeBench v677.1% 80% -2.9 ptGemma 4 31B
Only Gemma 4 26B A4B reports · 0
Gemma 4 26B A4B reports nothing Gemma 4 31B does not.
Only Gemma 4 31B reports · 0
Gemma 4 31B reports nothing Gemma 4 26B A4B does not.
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Gemma 4 26B A4B minus Gemma 4 31B in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecGemma 4 26B A4BGemma 4 31BEdge
Context window256K tokens256K tokensTie
Max output
Input price / 1M
Output price / 1M
Cached input / 1M
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
Modalitiestext · visiontext · visionTie
Released2026-04-022026-04-02
Cited benchmark scores66
Reliability

Google status

All providers
Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run Gemma 4 26B A4B and Gemma 4 31B on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.