Skip to content
← Models
ClosedSberbank/10 trustMetrics-checked

GigaChat 3.5 Ultra

Russia's largest bank-built model: 432B parameters, released July 2026.

Visit official page →

About

Released July 6, 2026, GigaChat 3.5 Ultra is the largest model Sberbank has ever released at 432 billion parameters, and the flagship of the GigaChat family — Russia's bank-built ChatGPT alternative, live since April 2023 with over 2.5 million users in its first ten months. It is built on Sber's NeONKA architecture (supervised fine-tuning plus RLHF, developed by SberDevices and Sber AI) and tuned specifically for Russian language, slang, and cultural context. Sberbank claims 4x faster long-text generation than the previous flagship. The GigaChat 2.0 generation (Pro and Max tiers) added direct audio understanding without speech-to-text conversion, emotion detection from intonation, and 200-page document handling.

Limitations: Russian-first training means English and other languages are secondary. Availability is concentrated in the Sber ecosystem and Russian market; international developers face API access friction. Independent benchmark verification is thin — most performance claims are vendor-supplied. In July 2026 Sberbank's own CEO publicly accused Yandex's competing YandexGPT-5 of being a fine-tuned Chinese Qwen, underscoring how contested Russian-model provenance claims are.

Specs

Context window

Not publicly disclosed (200 A4-page document support advertised)

Pricing

GigaChat API pricing (per-million-token, public tiers)

Free quota

Free tier via GigaChat assistant

API access

GigaChat API (api.giga.chat); GigaChat assistant

Weights

Closed — API only

Released

July 6, 2026

Benchmarks

Vendor-claimed: 4x faster long-text generation vs previous flagship; independent verification thin

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

Use cases

chatrussian-nlpdocument-analysisvoice

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 56; Trending score 70/100 from researcher assessment.

Sources