K2 Horizon 7B

Updated
15.09.2026
Thinking
Tools
Reasoning
Code
Run

K2 Horizon 7B from IFM: 512K context, dense weights and server setup. Atomic Chat support is not yet verified.

At a glance

  • License: Apache-2.0, per model card
  • Parameters: 7B core; 8.999B stored parameters
  • Context length: 524,288 tokens; native from midtraining
  • Modalities: Text input and output
  • Minimum hardware: Not measured; 18.00 GB BF16 files alone, plus runtime memory

What is K2 Horizon 7B?

K2 Horizon 7B is IFM's medium dense text model for coding, reasoning and tool-driven tasks. It has a native 512K context window from midtraining onward. You get a larger dense checkpoint than the 3.7B model, with published repository-repair and terminal-agent results to use as starting points for evaluation.

SpecificationK2 Horizon 7B
Parameters7B core; 8.999B stored parameters
ArchitectureDense decoder-only
Layers36
Context window524,288 tokens; native from midtraining
ModalitiesText input and output
Release statusReleased checkpoint
LicenseApache-2.0, per model card

The complete weight inventory is about 8.999B parameters, beyond the 7B core label. It uses 36 layers with hidden size 4,096. This page covers the base K2 Horizon 7B checkpoint, not a separately trained adapter or a community fine-tune; their capabilities and benchmark results must be checked independently.

K2 Horizon 7B benchmarks

IFM's source table changes reference-model order between rows. Here the five rows sharing Gemma 4-12B, Qwen3.5-9B and Granite 4.2-8B are aligned into fixed columns by model name. Scores are percentages reported by IFM, not Atomic Chat results. Source: official model card.

BenchmarkK2-Horizon-7BGemma 4-12BQwen3.5-9BGranite 4.2-8B
HMMT Feb 2026
Olympiad math
73.363.165.766.5
SWE-bench Verified
Software engineering
70.630.650.847.7
HLE
Expert questions
18.615.714.99.7
LCR
Long-context reasoning
68.061.765.343.3
Terminal-Bench 2.1
Terminal agents
39.127.329.218.4

K2 Horizon 7B leads all five displayed rows in the vendor comparison, including 70.6 on SWE-bench Verified and 39.1 on Terminal-Bench 2.1. This is a selected, named set of tests, not a universal model ranking. Tool availability, prompts and reasoning budget still affect an agent's success on your repository.

K2 Horizon 7B hardware requirements

The original BF16 weight files total 18.00 GB; IFM's BF16 GGUF totals 18.01 GB. GGUF packages the tokenizer and chat template for a supported llama.cpp loader. BF16 GGUF is not a 4-bit build and does not provide the memory reduction often associated with the format.

PrecisionSourceSize on disk
BF16 SafetensorsIFM/K2-Horizon-7B18.00 GB
BF16 GGUF (official)IFM/K2-Horizon-7B-GGUF
K2-Horizon-7B-BF16.gguf		18.01 GB

A 16 GB memory pool is smaller than the full BF16 weight set before any runtime allocations. Additional memory depends on the context you use and where the runtime places its cache. Review Qwen3.8-27B's published quantization options if you want to compare a model with an established lower-bit download ladder.

Sizes use decimal GB and cover weights only. For the format distinction, see what GGUF is. A memory tier is not listed because minimum runtime memory has not been measured for these builds.

How to run K2 Horizon 7B in Atomic Chat

Atomic Chat execution has not been verified for this checkpoint. IFM's official GGUF card requires a llama.cpp build with K2 Horizon architecture support and points to its development fork. Check architecture support before downloading these files. The usual search, download and start-chat flow is not yet a verified procedure for this model. Browse the model catalog for other checkpoints and their documented download options.

K2 Horizon 7B license

IFM labels this checkpoint Apache-2.0 in the official model card. Use the publisher's release information as the source for its license, and review any additional notices attached to the exact build you redistribute. A community conversion is not an Atomic Chat release.

Sources and file inventories checked September 15, 2026.

Get the weights from Hugging Face

# Download only; check disk space first.
hf download IFM/K2-Horizon-7B --revision ff325e226270e05ea081a97fd0c9c62652472fe8
For a compatible local server on port 8000:
curl http://127.0.0.1:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"IFM/K2-Horizon-7B","messages":[{"role":"user","content":"Write tests for a function that parses a CSV row."}],"temperature":1,"top_p":0.95,"max_tokens":32768,"chat_template_kwargs":{"reasoning_effort":"high"}}'
Install the OpenAI Python client and start a compatible local server first.
from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8000/v1", api_key="local")
result = client.chat.completions.create(
    model="IFM/K2-Horizon-7B",
    messages=[{"role": "user", "content": "Write tests for a CSV parser."}],
    temperature=1.0, top_p=0.95, max_tokens=32768,
    extra_body={"chat_template_kwargs": {"reasoning_effort": "high"}},
)
print(result.choices[0].message.content)
Requires a compatible local server and the OpenAI JavaScript client.
import OpenAI from "openai";
const client = new OpenAI({baseURL: "http://127.0.0.1:8000/v1", apiKey: "local"});
const result = await client.chat.completions.create({
  "model": "IFM/K2-Horizon-7B",
  "messages": [
    {
      "role": "user",
      "content": "Write tests for a function that parses a CSV row."
    }
  ],
  "temperature": 1,
  "top_p": 0.95,
  "max_tokens": 32768,
  "chat_template_kwargs": {
    "reasoning_effort": "high"
  }
});
console.log(result.choices[0].message.content);
Desktop
macOS
(Intel and Apple Silicon)
Download
Windows
(x64)
Download
Linux
(x86_64)
Download

Frequently asked questions

K2 Horizon 7B is IFM's medium dense text model for coding, reasoning and tool-driven tasks. It has a native 512K context window from midtraining onward. You get a larger dense checkpoint than the 3.7B model, with published repository-repair and terminal-agent results to use as starting points for evaluation.

Execution of this checkpoint has not been verified in Atomic Chat. The official GGUF requires a llama.cpp build with K2 Horizon support. A downloadable GGUF does not by itself confirm that the app can load it.

Its original BF16 weight files total 18.00 GB. That is disk usage, not a tested minimum RAM or VRAM figure. You also need memory for the inference engine and a cache that grows with context.

This page describes IFM/K2-Horizon-7B, the released dense checkpoint with 512K context. An adapter or fine-tune is a separate artifact. Its requirements and results should not be substituted for the base model's figures.

IFM labels this checkpoint Apache-2.0 in the official model card. Use the publisher's release information as the source for its license, and review any additional notices attached to the exact build you redistribute. A community conversion is not an Atomic Chat release.