Sber and Yandex Cut LLM API Prices in Domestic Token Price War

Russian technology briefings in early-to-mid September 2026 described a sharp domestic price contest between Sber’s GigaChat and Yandex language models. Reports said GigaChat Lite input tokens fell on the order of two-thirds to about 0.065 rubles per thousand, while YandexGPT Pro rates dropped roughly one-third, with larger enterprise discounts available above high monthly volumes.

Filed under Business and dated September 15, 2026, this AI4Russia briefing treats the cuts as Russian cloud-and-API market news—not a quality ranking. The moves arrive as Yandex continues unifying generative products under the Alice AI brand and as Sber promotes reasoning-capable open weights for developers.

Why it matters: Russian product teams deciding between domestic APIs care about unit economics as much as benchmarks. Cheaper tokens can unlock higher-volume customer-service and document pipelines—but only if latency, logging, and data-residency terms stay explicit in contracts.

What it means in practice

Sber and Yandex Cut LLM API Prices in Domestic Token Price War — contextual photo

Russian engineering leads should inventory which workloads can move to Lite or Flash tiers; confirm SLA and logging clauses after price changes; assign an owner for prompt-cost monitoring; run time-boxed A/B tests on quality versus spend; and prefer dual-vendor fallbacks. Place pricing shifts beside Yandex’s Alice AI brand unification and Sber’s GigaChat reasoning release.

Caveats come first. List prices change; enterprise quotes differ; and cheaper models may not match flagship quality on complex tasks. AI4Russia therefore presents the September cuts as directional commercial context until audited production benchmarks are published by buyers.

What to watch next: Alice AI Studio rate cards after API renames; whether open-weight hosting undercuts managed APIs further; and published case studies from banks and retailers. Readers following draft sovereign-model rules can continue on the AI4Russia homepage, or browse the Newsroom for additional briefings.

Bottom line: treat this update as orientation, not instruction. Russian LLM pricing is competitive, fluid, and still early for many SMEs. Organizations that benefit most will measure quality per ruble, keep audit logs, and refuse to confuse a promotional rate card with production readiness.