Synthszr Charts — die großen AI-Marken im Wettkampf ums Podium
synthszr charts
google

Gemma 4

#1

google · v4 · seit 2026-04-02 · 319× · zuletzt 30. Juni 2026

100
Momentum

Google Gemma 4 ist eine von Google DeepMind am 2. April 2026 veröffentlichte Open-Weight-Modellfamilie, lizenziert unter Apache 2.0. Sie umfasst fünf Größen (E2B, E4B, 12B, 26B A4B, 31B) mit zwei Architekturen – dicht (31B) und Mixture-of-Experts (26B A4B) – und unterstützt als erste Gemma-Generation Text, Bild, Audio und Video nativ über alle Größen hinweg. Die Modelle sind sowohl für On-Device-Einsatz (Smartphones, Edge) als auch für Consumer-GPUs und Workstations ausgelegt und enthalten konfigurierbare Reasoning-Modi (Thinking Mode) sowie nativen Function Calling.

Momentum-Verlauf
04.04.03.07.

Features

Key-Benchmark (%)AIME 2026: 89,2 % (31B); GPQA Diamond: 84,3 % (31B); LMArena Score: 1452 (31B) / 1441 (26B MoE)
Kontextfenster (Token)128K (E2B, E4B); 256K (12B, 26B A4B, 31B)
LizenzApache 2.0 – kommerzielle Nutzung ohne Einschränkungen (keine MAU-Caps)
MultimodalitätInput: Text + Bild (alle Varianten), Video & Audio nativ (E2B, E4B, 12B); Output: Text only
PlattformHugging Face, Kaggle, Ollama (Weights); Google AI Studio, Vertex AI (API); Lokal: llama.cpp, vLLM, MLX, LM Studio, Ollama; On-Device: Android AICore, LiteRT-LM
PreisGewichte kostenlos (Open-Weight); API via Google AI Studio / Drittanbieter (siehe Preis pro 1M Token)
Preis pro 1M Token31B: $0,12 Input / $0,35 Output; 26B A4B: $0,06 Input / $0,30–0,33 Output; E4B: $0,20/$0,20; E2B: kostenlos (via Google)
Release-Datum2. April 2026 (E2B/E4B/26B/31B); 12B Unified folgte später (Mai/Juni 2026)

Belege (60)

Subscribe free. Unsubscribe the second it sucks.

High-signal news across AI, business, UX, and tech. Every morning.

Gemma 4 — Synthszr Ranking