ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

Gemma 4 E2BvsGemma 4 E4B

Gemma 4 E4B leads 4 of 4 shared benchmarks. Both offer a 128K-token context window.

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

Shared benchmarks
0 – 4
Gemma 4 E4B leads
Cheaper per token
list price, input + output
Larger context
Tie
both 128K tokens
Provider uptime (30d)
100%
Google
Gemma 4 E2B
Google · released 2026-04-02
textvisionaudio
Context
128K
Max out
Input /1M
Output /1M
Cached /1M
Scores
4 · 3 core
Gemma 4 E4B
Google · released 2026-04-02
textvisionaudio
Context
128K
Max out
Input /1M
Output /1M
Cached /1M
Scores
4 · 3 core
Quality

Benchmark matrix

BenchmarkGemma 4 E2BGemma 4 E4BΔEdge
Reported by both · 4
GPQA Diamond43.4% 58.6% -15.2 ptGemma 4 E4B
MMLU-Pro60% 69.4% -9.4 ptGemma 4 E4B
AIME 202637.5% 42.5% -5 ptGemma 4 E4B
LiveCodeBench v644% 52% -8 ptGemma 4 E4B
Only Gemma 4 E2B reports · 0
Gemma 4 E2B reports nothing Gemma 4 E4B does not.
Only Gemma 4 E4B reports · 0
Gemma 4 E4B reports nothing Gemma 4 E2B 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 E2B minus Gemma 4 E4B 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 E2BGemma 4 E4BEdge
Context window128K tokens128K 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 · vision · audiotext · vision · audioTie
Released2026-04-022026-04-02
Cited benchmark scores44
Reliability

Google status

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

Public benchmarks are a starting point. Run Gemma 4 E2B and Gemma 4 E4B on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.