Capabilities · Concept
Gemma 4: Open-Weight Natively Multimodal Language Models
Open-weight Google model family used as frozen base for lightweight safe error-correction modules.
Forecast horizon
Sets new open-weight baseline for multimodal reasoning; likely to accelerate open-source agentic deployments.
Connections
Connections · 2
How this node ties into the rest of the map, and the evidence behind each link.
from · publishes 4/10
Google DeepMind Google DeepMind introduced Gemma 4, a new generation of open-weight natively multimodal language models.
+4 growth to · applies to 3/10
Science of AI Evaluation Gemma 4 establishes new performance benchmarks across STEM, multimodal, and long-context tasks, advancing evaluation science.
+3 growthSignal sources
Signal sources
Dated facts from primary sources in this direction.
Task horizon doubling Mar 2025
The length of software tasks AI agents can do autonomously at 50% reliability has doubled about every 7 months — and since 2024 closer to every ~3 months.
METR → Benchmarks saturating Apr 2025
In one year scores rose by 18.8, 48.9 and 67.3 points on MMMU, GPQA and SWE-bench; real-world software solve rate jumped from 4.4% to 71.7%.
Stanford HAI — AI Index 2025 → Autonomous coding 2025–2026
On SWE-bench Verified (500 real GitHub issues), autonomous coding agents reached ~80–86% by late 2025, up from under 50% in early 2025.
Epoch AI →