H2O-Danube3-4B
Efficient 4B small language model for chat and on-device use.
About
H2O-Danube3-4B is a 4B-parameter small language model from H2O.ai trained on 6 trillion tokens and released under Apache 2.0. It scores over 80% on the 10-shot HellaSwag benchmark, which H2O.ai says beats Apple's OpenELM-3B and rivals Microsoft's Phi-3 4B.
It is built for economical, edge-friendly deployment — chat, research, and offline applications — though as a 2024-era 4B model it will not match current frontier models on hard reasoning.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
80%+ on 10-shot HellaSwag (H2O.ai-reported).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Apache 2.0, 6T training tokens, edge-friendly positioning; 2024 release; older generation vs current small models; HellaSwag leadership claim is company-reported. License: Apache 2.0 per H2O.ai's open-source release..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Apache 2.0, 6T training tokens, edge-friendly positioning
- · 2024 release; older generation vs current small models
- · HellaSwag leadership claim is company-reported