DedeProGames/orion:2b

DedeProGames/ orion:2b

76 Downloads Updated 1 month ago

Orion is an open-source language model built for strong general reasoning and reliable text generation, with agentic capabilities for on device workflows.

vision tools thinking 2b 30b

ollama run DedeProGames/orion:2b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "DedeProGames/orion:2b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='DedeProGames/orion:2b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'DedeProGames/orion:2b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 month ago

1 month ago

324d162be6ca · 2.7GB ·

model

archqwen35

parameters2.27B

quantizationQ8_0

2.7GB

license

Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US

11kB

params

{ "presence_penalty": 1.5, "temperature": 1, "top_k": 20, "top_p": 0.95 }

65B

Readme

Orion is an open-source programming language model developed for robust general reasoning and reliable text generation, with agent-based capabilities for cross-device workflows. Available in 2 sizes.

DedeProGames/orion:30b

Base: GLM-4.7-Flash
Goal: Best overall reasoning quality in the Orion family
When to use: Math, logic, code reasoning, multi-step tasks, long-form answers

DedeProGames/orion:2b

Base: Qwen3.5 (2b Class)
Objective: Powerful reasoning for complex tasks that can be executed on weaker devices.
When to use: Fast iteration, assistants on older PCs/laptops, smaller VRAM setups

Benchmarks

Benchmark	orion	gpt-oss:20b	qwen3.5:27b	Astral-2.6
AIME 25	91.6	91.7	—	—
GPQA	75.2	71.5	85.5	83.8
LCB v6	64.0	*61.0	80.7	76.9
HLE	14.4	10.9	24.3	12.6
SWE-bench Verified	59.2	60.7	72.4	—
τ²-Bench	79.5	*47.7	79.0	48.57
BrowseComp	42.8	*28.3	61.0	19.12

![Orion-Image.png](/assets/DedeProGames/orion/840b148f-ba37-40ed-b654-a7c4ae6057a7)

Orion is an **open-source** programming language model developed for **robust general reasoning** and reliable text generation, with **agent-based capabilities** for **cross-device workflows**. Available in 2 sizes.

---

### DedeProGames/orion:30b
- Base: GLM-4.7-Flash
- Goal: Best overall reasoning quality in the Orion family
- When to use: Math, logic, code reasoning, multi-step tasks, long-form answers

### DedeProGames/orion:2b
- Base: Qwen3.5 (2b Class)
- Objective: Powerful reasoning for complex tasks that can be executed on weaker devices.
- When to use: Fast iteration, assistants on older PCs/laptops, smaller VRAM setups

---

## Benchmarks

| Benchmark            | orion | gpt-oss:20b | qwen3.5:27b | Astral-2.6 |
|---------------------|---------------|-------------|-------------|------------|
| AIME 25             | 91.6          | 91.7    | —           | —          |
| GPQA                | 75.2          | 71.5        | 85.5    | 83.8       |
| LCB v6              | 64.0          | *61.0       | 80.7    | 76.9       |
| HLE                 | 14.4          | 10.9        | 24.3    | 12.6       |
| SWE-bench Verified  | 59.2          | 60.7        | 72.4    | —          |
| τ²-Bench            | 79.5      | *47.7       | 79.0        | 48.57      |
| BrowseComp          | 42.8          | *28.3       | 61.0    | 19.12      |

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)