Salamandra-7B-Instruct
Multilingual 7B trained from scratch on 35 European languages.
About
The Barcelona Supercomputing Center's language model, trained from scratch on 35 European languages plus code — 12.875 trillion tokens of curated open-access data.
The instruct version is a proof of concept rather than a polished product (no RLHF alignment), but the Apache 2.0 license and open training scripts make it a strong base for European-language work.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Competitive with similarly-sized open models on multilingual benchmarks per the Salamandra technical report.
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 (metrics tier, metrics-v1). Evidence: Niche adoption centered on European-language NLP research; BSC's Language Technologies Unit maintains the family; instruct version is a proof of concept without RLHF; Apache 2.0 weights, training scripts, and configs all public. Manual review: 35-language coverage and 12.875T token figure taken from the technical report. License: Apache 2.0.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 5.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Niche adoption centered on European-language NLP research
- · BSC's Language Technologies Unit maintains the family; instruct version is a proof of concept without RLHF
- · Apache 2.0 weights, training scripts, and configs all public
- · manual review: 35-language coverage and 12.875T token figure taken from the technical report