Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Sber / GigaChat 2

GigaChat 2 Pro

Multimodal model for complex instructions, summarization, editing and question answering.

Language models (LLMs)
availableClosed weightsText generation
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
10 of 17 specification fields collected
Context window128K128,000 tokens
Maximum outputNot collectedAwaiting source review
Input / 1M tokensRUB 500.00GigaChat API
Output / 1M tokensRUB 500.00RUB · billing conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
128,000 tokensSource for Context / input limit (opens in a new tab)

Provider-published context window, not an independently measured maximum output allowance.

Maximum input tokens
Not collected
Maximum output
Not collected
Reasoning
Not collected
Tool calling
SupportedSource for Tool calling (opens in a new tab)

Custom function argument generation; execution is performed by the caller. Built-in functions are separate hosted tools.

Structured output
SupportedSource for Structured output (opens in a new tab)

JSON Schema or regular-expression constrained generation; provider guide applies to GigaChat and demonstrates GigaChat-2-Max.

Streaming
SupportedSource for Streaming (opens in a new tab)

stream=true via server-sent events on the chat endpoint.

Batch processing
SupportedSource for Batch processing (opens in a new tab)

Pay-as-you-go only; asynchronous JSONL processing. Built-in functions are unavailable in batch mode.

Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
2025-03-13Source for Release date (opens in a new tab)

API release date of the GigaChat 2 family; served revisions have subsequently changed.

Parameters
Not collected
Languages
Not collected
Context caching
SupportedSource for Context caching (opens in a new tab)

Session-based shared-prefix reuse; hit tokens are not billed.

Input modalities
Text, Image, AudioSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)

Image and 3D generation use built-in external functions; they are not asserted native output modalities.

Billing basis
RUB including VAT. Existing SaluteDevices / Salute for Business customers on pay-as-you-go. From 1 September 2026 new customers and customers without prior package purchases are directed to Cloud.ru, whose prices are not represented here. Minimum total monthly bill is RUB 600 when usage is nonzero; zero usage has no charge. These unit costs exclude that account-level minimum and are not a complete bill.Source for Billing basis (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

GigaChat API

Provider source (opens in a new tab)
API model ID
GigaChat-2-Pro

Legacy alias: GigaChat-Pro.

Access
POST https://api.giga.chat/v1/chat/completions; authorized access token required.
Regions
Processing region and residency contract not established by this review.
Concurrent requests
Individuals: 1 concurrent request. Businesses: 10 by default; increases by arrangement.Source for Concurrent requests (opens in a new tab)

Account-level default, not a separate allowance for each model; batch throughput is not inferred from it.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Standard business usageInput1RUB 500.001M tokens
Standard business usageOutput1RUB 500.001M tokens
Standard business usageCached input1RUB 0.001M tokens

Rate conditions

  1. RUB including VAT. Existing SaluteDevices / Salute for Business customers on pay-as-you-go. From 1 September 2026 new customers and customers without prior package purchases are directed to Cloud.ru, whose prices are not represented here. Minimum total monthly bill is RUB 600 when usage is nonzero; zero usage has no charge. These unit costs exclude that account-level minimum and are not a complete bill. Uncached input and generated output tokens use the same rate; cache hits are subtracted from the billable token total.
Pricing source (opens in a new tab)Pricing source (opens in a new tab)

Data & enterprise terms

Retention and training-use commitments not verified. A cache mechanism is not a no-logging guarantee.

GigaChat API

Provider source (opens in a new tab)
API model ID
GigaChat-2-Pro

Legacy alias: GigaChat-Pro.

Access
Asynchronous JSONL batch API; pay-as-you-go account required. Submit requests through the provider batch workflow.
Regions
Processing region and residency contract not established by this review.
Batch processing
SupportedSource for Batch processing (opens in a new tab)

Built-in functions unavailable; custom function arguments can be generated.

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Asynchronous business usageInput1RUB 250.001M tokens
Asynchronous business usageOutput1RUB 250.001M tokens

Rate conditions

  1. RUB including VAT. Existing SaluteDevices / Salute for Business customers on pay-as-you-go. From 1 September 2026 new customers and customers without prior package purchases are directed to Cloud.ru, whose prices are not represented here. Minimum total monthly bill is RUB 600 when usage is nonzero; zero usage has no charge. These unit costs exclude that account-level minimum and are not a complete bill. Uncached input and generated output tokens use the same rate; cache hits are subtracted from the billable token total.
Pricing source (opens in a new tab)

Data & enterprise terms

Retention and training-use commitments not verified. A cache mechanism is not a no-logging guarantee.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

GigaChat API

GigaChat-2-Pro
Documented
Support evidence (opens in a new tab)
How it works
Session-based prefix cachingSource for How it works (opens in a new tab)
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Reuse X-Session-ID and matching conversation context. If omitted, the server assigns a UUID4 session ID.Source for How to enable it (opens in a new tab)
Usage fields
Inspect usage.precached_prompt_tokens; these tokens are excluded from billable total_tokens.Source for Usage fields (opens in a new tab)
Billing rules
Cache-hit tokens are not charged; uncached prompt and generated tokens are billed.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Requires the same session identifier and partially matching context. Cache hits are not guaranteed; no fixed lifetime or minimum token threshold verified.Source for Limits & data boundaries (opens in a new tab)
Cache reads, writes & storage
ChargeApplies toRate / unit
Cached input1Standard business usageRUB 0.00 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. RUB including VAT. Existing SaluteDevices / Salute for Business customers on pay-as-you-go. From 1 September 2026 new customers and customers without prior package purchases are directed to Cloud.ru, whose prices are not represented here. Minimum total monthly bill is RUB 600 when usage is nonzero; zero usage has no charge. These unit costs exclude that account-level minimum and are not a complete bill. Uncached input and generated output tokens use the same rate; cache hits are subtracted from the billable token total.

Observed 14 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 14 Oct 2026

GigaChat API

GigaChat-2-Pro
Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 14 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Exact native hosted selector documented by Sber. No matching downloadable checkpoint was identified in the reviewed public ai-sage artifact inventory. Public GigaChat 20B, GigaChat 3.x and Giga embedding releases are not assumed identical to this hosted model; this is a qualified catalog classification, not proof about undisclosed private variants.

Classification evidence (opens in a new tab)

Lifecycle dates

GigaChat 2 family introduced; initial 27.4 revision is now unavailable13 Mar 2025Source (opens in a new tab)
28.2 revision introduced FP8/W8A16 inference changes15 Apr 2025Source (opens in a new tab)
30.1 family revision; Max text generation improved23 Sept 2025Source (opens in a new tab)

Sources reviewed

Sber max (opens in a new tab)Reviewed 14 Oct 2026
Sber pro (opens in a new tab)Reviewed 14 Oct 2026
Sber lite (opens in a new tab)Reviewed 14 Oct 2026