Model library / yandex
AliceAI-Foundation-80B-A3B-Base is an 81-billion-parameter autoregressive language model with mixture-of-experts layers, trained from scratch for Russian factual and reasoning tasks.
Mixture of experts, count not published
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
The model is a causal language model released by Yandex. It uses a hybrid architecture with MoE layers, 48 layers total, and is intended for complex reasoning, tool use and Russian-language factual knowledge tasks, as described by the publisher.
It contains 81.3 billion parameters, of which 3 billion are active per token, and supports a context window of 262 144 tokens. The model is licensed under Apache-2.0, includes 512 experts with a top-10 routing plus a shared expert, and its weight format is not published.
Specification
| Repository | yandex/AliceAI-Foundation-80B-A3B-Base |
| Publisher | yandex |
| Published | 12 September 2026 |
| Architecture | AliceAIForCausalLM |
| Model type | alice_ai |
| Parameters | 81,286,433,408 |
| Layers | 48 |
| Hidden size | 2,048 |
| Attention heads | 16, 2 key/value heads |
| Context length | 262,144 tokens |
| Vocabulary | 129,024 tokens |
| Licence | apache-2.0 |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 151.4 GiB | Full precision as released |
| fp8 | 75.7 GiB | 8-bit, near-lossless on most models |
| int4 | 37.9 GiB | 4-bit, smallest footprint |
The cache costs 96 KiB per token at 16-bit, so the full 262,144-token context of one request needs about 24.0 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 8 | 1.4 GiB | about 15,206 tokens |
| RTX 5090 32GB | 8 | 59.0 GiB | the full 262,144 tokens |
| L40S 48GB | 4 | 11.4 GiB | about 124,433 tokens |
| A100 80GB | 4 | 126.6 GiB | the full 262,144 tokens |
| H100 80GB | 4 | 126.6 GiB | the full 262,144 tokens |
| RTX PRO 6000 96GB | 2 | 16.4 GiB | about 179,046 tokens |
| H200 141GB | 2 | 97.4 GiB | the full 262,144 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve yandex/AliceAI-Foundation-80B-A3B-Base \ --tensor-parallel-size 8 \ --max-model-len 12288 \ --gpu-memory-utilization 0.90
Questions
The weights take about 151 GiB at the released precision. No single card in the table above holds that, so it needs several GPUs or a lower precision.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 262,144 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
Ling-3.0-flash-Fin is a 127.5 billion-parameter finance-enhanced language model for long-context financial research and analysis.
Ling-3.0-flash is a 127.5 B parameter hybrid-linear language model from inclusionAI for reasoning and long-context applications.
Laguna S 2.1 is a 118 billion-parameter mixture-of-experts model for agentic coding with long-context support.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.