Model library / IFM
A 0.9B decision model that returns probabilities for multiple questions in a single forward pass, not a text generator.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
The IFM K2-Type-0.9B model is a decision-making system built on the K2-Horizon-0.9B backbone. It accepts a state (text or JSON) and any number of typed questions, then outputs a probability for each option in one forward pass. The model uses a pointer head and a specialised input layout, and it is intended to be served via the provided HTTP server using TypeSafe's /v1/systemone wire format.
The model stores 1.1 billion parameters in bfloat16 format and supports a context length of 131 072 tokens. It is released under the Apache-2.0 licence. Deployment requires transformers ≥ 5.17 and torch 2.8, with typical inference latency of 20-60 ms after CUDA warm-up. The backbone is not suitable for text generation; only the decision head should be used.
Specification
| Repository | IFM/K2-Type-0.9B |
| Publisher | IFM |
| Published | 26 September 2026 |
| Architecture | K2HorizonForCausalLM |
| Model type | k2_horizon |
| Parameters | 1,078,285,824 |
| Layers | 28 |
| Hidden size | 1,536 |
| Attention heads | 32, 8 key/value heads |
| Context length | 131,072 tokens |
| Vocabulary | 64,256 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
| Base model | IFM/K2-Horizon-0.9B |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 2.0 GiB | Full precision as released |
| fp8 | 1.0 GiB | 8-bit, near-lossless on most models |
| int4 | 0.5 GiB | 4-bit, smallest footprint |
The cache costs 56 KiB per token at 16-bit, so the full 131,072-token context of one request needs about 7.0 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 | 1 | 17.1 GiB | the full 131,072 tokens |
| RTX 5090 32GB | 1 | 24.3 GiB | the full 131,072 tokens |
| L40S 48GB | 1 | 38.7 GiB | the full 131,072 tokens |
| A100 80GB | 1 | 67.5 GiB | the full 131,072 tokens |
| H100 80GB | 1 | 67.5 GiB | the full 131,072 tokens |
| RTX PRO 6000 96GB | 1 | 81.9 GiB | the full 131,072 tokens |
| H200 141GB | 1 | 122.4 GiB | the full 131,072 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve IFM/K2-Type-0.9B \ --max-model-len 131072 \ --gpu-memory-utilization 0.90
Questions
The weights take about 2 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 17 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 131,072 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
K2-Horizon-375B-A23B is an open-weight sparse MiE language model from IFM with 379 B parameters, 23 B active per token and a 512 …
K2-Horizon-7B is a 9-billion-parameter decoder-only language model with a 524k token context window, released under Apache-2.0.
K2-Horizon-3.7B is a 3.7 billion-parameter dense decoder-only model with a 512 K token context, released openly by IFM.
K2-Horizon-32B-Stage1 is a 32-billion-parameter dense decoder-only model released by IFM for benchmarking agentic, coding and rea…
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.