Skip to content
← Models
Open weightsIlya Gusev/10 trustMetrics-checked

Saiga

Community-standard Russian instruction-tuned chat models on Llama and Mistral bases.

Visit official page →

About

Saiga is a family of Russian-language instruction-tuned models maintained by Ilya Gusev, among the most widely used open Russian chat models. Builds such as saiga_llama3_8b start from Meta's Llama 3 8B Instruct and add several rounds of supervised fine-tuning and preference optimization (KTO) on Russian datasets including the curated saiga_scored collection.

The models use a Russian system prompt identity ("Sayga, Russian-language automatic assistant") and are a staple baseline in Russian LLM benchmarks and community fine-tuning work. Licensing follows the base model (Llama 3 community license for the Llama-based builds), so commercial use inherits Meta's terms.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

June 1, 2023

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

russian-chatrussian-instruction-followingregional-llm-baseline

Trust score — 6/10

/10 trustMetrics-checked

Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 109; Trending score 55/100 from researcher assessment.

Sources