Model library / Aleph-Alpha
Kolibri is a 78.1 billion-parameter mixture-of-experts language model from Aleph-Alpha, supporting German and English with long-context and tool-calling capabilities.
Mixture of experts, count not published
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Kolibri is Aleph-Alpha's mixture-of-experts reasoning model aimed at German and English text processing. The publisher states it is intended for multi-step reasoning, coding, structured extraction, retrieval-augmented generation, long-document handling and agentic tool calling, with an explicit reasoning mode and tool-calling support.
The model comprises 78.1 billion parameters in a MoE architecture that activates only a subset per token, uses fp8 weights, and is released under the Apache-2.0 licence. Its native context length is 262,144 tokens, extendable up to 1,048,576 tokens, and it requires the aleph-alpha-inference package for deployment.
Specification
| Repository | Aleph-Alpha/Kolibri-1 |
| Publisher | Aleph-Alpha |
| Published | 2 October 2026 |
| Architecture | Kolibri1ForCausalLM |
| Model type | kolibri1 |
| Parameters | 78,103,074,560 |
| Layers | 50 |
| Hidden size | 2,560 |
| Attention heads | 48, 4 key/value heads |
| Context length | 262,144 tokens |
| Vocabulary | 128,000 tokens |
| Quantisation | fp8 |
| Licence | apache-2.0 |
| Base model | Aleph-Alpha/Kolibri-1-BF16 |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 145.5 GiB | Full precision as released |
| fp8 | 72.7 GiB | 8-bit, near-lossless on most models |
| int4 | 36.4 GiB | 4-bit, smallest footprint |
The cache costs 100 KiB per token at 16-bit, so the full 262,144-token context of one request needs about 25.0 GiB. Concurrency multiplies that number, not the weights.
That figure is an upper bound: this model uses sliding-window attention with a 513-token window on some layers, so long requests cache less than this.
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 | 4 | 3.0 GiB | about 31,501 tokens |
| RTX 5090 32GB | 4 | 31.8 GiB | the full 262,144 tokens |
| L40S 48GB | 2 | 8.0 GiB | about 83,930 tokens |
| A100 80GB | 2 | 65.6 GiB | the full 262,144 tokens |
| H100 80GB | 2 | 65.6 GiB | the full 262,144 tokens |
| RTX PRO 6000 96GB | 1 | 10.5 GiB | about 110,144 tokens |
| H200 141GB | 1 | 51.0 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 Aleph-Alpha/Kolibri-1 \ --max-model-len 28672 \ --gpu-memory-utilization 0.90
Questions
The weights take about 73 GiB at the released precision. That fits on one RTX PRO 6000 96GB, which leaves roughly 11 GiB for the key-value cache.
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 hello@aiqre.com with the model name and roughly what volume you expect.
Also in the library
Computer-10 is a 70-billion-parameter Llama 3.1-based document model that generates conversational turns without instruction tuni…
AliceAI-Foundation-80B-A3B-Base is an 81-billion-parameter autoregressive language model with mixture-of-experts layers, trained …
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.
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.